protocolfield.cpp

#include "protocolfield.h"
#include "protocolparser.h"
#include "shuntingyard.h"
#include "enumcreator.h"
#include "protocolstructuremodule.h"
#include "protocolbitfield.h"
#include "prebuiltSources/floatspecial.h"
#include <cmath>
#include <iomanip>
#include <sstream>
#include <regex>
#include <limits>

TypeData::TypeData(ProtocolSupport sup) :
    isBool(false),
    isStruct(false),
    isSigned(false),
    isBitfield(false),
    isFloat(false),
    isEnum(false),
    isString(false),
    isFixedString(false),
    isNull(false),
    bits(8),
    sigbits(0),
    enummax(0),
    support(sup)
{
}


TypeData::TypeData(const TypeData& that) :
    isBool(that.isBool),
    isStruct(that.isStruct),
    isSigned(that.isSigned),
    isBitfield(that.isBitfield),
    isFloat(that.isFloat),
    isEnum(that.isEnum),
    isString(that.isString),
    isFixedString(that.isFixedString),
    isNull(that.isNull),
    bits(that.bits),
    sigbits(that.sigbits),
    enummax(that.enummax),
    enumName(that.enumName),
    support(that.support)
{
}


/*!
 * Reset all members to defaults except the protocol support
 */
void TypeData::clear(void)
{
    isBool = false;
    isStruct = false;
    isSigned = false;
    isBitfield = false;
    isFloat = false;
    isEnum = false;
    isString = false;
    isFixedString = false;
    isNull = false;
    bits = 8;
    sigbits = 0;
    enummax = 0;
    enumName.clear();
}


/*!
 * Determine the typename of this field (for example uint8_t).
 * \param structName is the structure name, if this is a structure.
 * \return the type name that should appear in code.
 */
std::string TypeData::toTypeString(const std::string& structName) const
{
    std::string typeName;

    if(isString)
        typeName = "char";
    else if(isBitfield)
    {
        if((bits > 32) && (support.longbitfield))
            typeName = "uint64_t";
        else
            typeName = "unsigned";
    }
    else if(isEnum)
    {
        typeName = enumName;
    }
    else if(isStruct)
    {
        typeName = trimm(structName);

        // Make sure it ends with the suffix;
        if(!endsWith(typeName, support.typeSuffix))
        {
            typeName += support.typeSuffix;
        }
    }
    else
    {
        // create the type name
        if(isBool)
            typeName = "bool";
        else if(isFloat)
        {
            if(bits > 32)
                typeName = "double";
            else
                typeName = "float";

        }// if floating point type
        else
        {
            if(isSigned)
                typeName = "int";
            else
                typeName = "uint";

            // Add the bits, we use only valid native type widths
            if(bits > 32)
                typeName += "64_t";
            else if(bits > 16)
                typeName += "32_t";
            else if(bits > 8)
                typeName += "16_t";
            else
                typeName += "8_t";

        }// else if integer type
    }

    return typeName;

}// toTypeString


/*!
 * Determine the signature of this field (for example uint8).
 * \return the signature name of this field.
 */
std::string TypeData::toSigString(void) const
{
    if(isString || isBitfield || isStruct || isBool)
        return "unknown";
    else if(isFloat)
    {
        return "float" + std::to_string(bits);

    }// if floating point type
    else
    {
        if(isSigned)
            return "int" + std::to_string(bits);
        else
            return "uint" + std::to_string(bits);

    }// else if integer type

}// toSigString


/*!
 * Determine the maximum floating point value this TypeData can hold
 * \return the maximum floating point value of this TypeData
 */
double TypeData::getMaximumFloatValue(void) const
{
    if(isString || isStruct)
        return 0;
    else if(isEnum)
    {
        return enummax;
    }
    else if(isFloat)
    {
        // Float encodings use float rules
        if(bits <= 16)
        {
            return float16ToFloat32(float32ToFloat16( std::numeric_limits<float>::max(), sigbits), sigbits);
        }
        else if(bits <= 24)
        {
            return float24ToFloat32(float32ToFloat24( std::numeric_limits<float>::max(), sigbits), sigbits);
        }
        else if(bits <= 32)
        {
            return std::numeric_limits<float>::max();
        }
        else
            return std::numeric_limits<double>::max();
    }
    else
    {
        return (double)getMaximumIntegerValue();
    }

}// TypeData::getMaximumFloatValue


/*!
 * Determine the minimum floating point value this TypeData can hold
 * \return the minimum floating point value of this TypeData
 */
double TypeData::getMinimumFloatValue(void) const
{
    if(isString || isStruct)
        return 0;
    else if(isFloat)
    {
        // Float encodings use float rules
        if(bits <= 16)
        {
            // Note that min() is the smallest positive number
            return float16ToFloat32(float32ToFloat16(-1*std::numeric_limits<float>::max(), sigbits), sigbits);
        }
        else if(bits <= 24)
        {
            return float24ToFloat32(float32ToFloat24(-1*std::numeric_limits<float>::max(), sigbits), sigbits);
        }
        else if(bits <= 32)
        {
            return -1*std::numeric_limits<float>::max();
        }
        else
            return -1*std::numeric_limits<double>::max();
    }
    else
    {
        return (double)getMinimumIntegerValue();
    }

}// TypeData::getMinimumFloatValue


/*!
 * Determine the maximum integer value this TypeData can hold
 * \return the maximum integer value of this TypeData
 */
uint64_t TypeData::getMaximumIntegerValue(void) const
{
    if(isString || isStruct)
        return 0;
    else if(isEnum)
    {
        return enummax;
    }
    else if(isFloat)
    {
        return (uint64_t)getMaximumFloatValue();
    }
    else if(isSigned)
    {
        uint64_t max = (0x1ull << (bits-1)) - 1;
        return max;
    }
    else
    {
        uint64_t max = (0x1ull << bits) - 1;
        return max;
    }

}// TypeData::getMaximumIntegerValue


/*!
 * Determine the minimum integer value this TypeData can hold
 * \return the minimum integer value of this TypeData
 */
int64_t TypeData::getMinimumIntegerValue(void) const
{
    if(isString || isStruct)
        return 0;
    else if(isFloat)
    {
        return (int64_t)getMinimumFloatValue();
    }
    else if(isSigned)
    {
        // We have to deal with the sign extension here. For example -128 for an
        // 8-bit number is 0x80, but -128 for a 16-bit number is 0xFF80. Since we
        // are returning a 64-bit number it is almost always bigger than the
        // number of bits and sign extensions must be applied
        int64_t min = (0xFFFFFFFFFFFFFFFFll << (bits-1));
        return min;
    }
    else
    {
        return 0;
    }

}// TypeData::getMinimumIntegerValue


/*!
 * Given a constant string (like default value) apply the type correct suffix
 * for the type, or apply a cast if a suffix cannot be used.
 * \param number is the input number whose type must be set.
 * \return The correctly typed number, using either a constant suffix or a cast.
 */
std::string TypeData::applyTypeToConstant(const std::string& number) const
{
    if(isString || isStruct)
        return number;

    std::string output = trimm(number);

    if(output.empty())
    {
        if(isEnum && !enumName.empty())
            return "(" + enumName + ")0";
        else
            return "0";
    }

    // Remove the existing suffix
    if(startsWith(number, "0b"))
    {
        while(!output.empty() && !ShuntingYard::isNumber(output.back(), false, true))
            output.erase(output.size() - 1, 1);
    }
    else if(startsWith(number, "0x"))
    {
        while(!output.empty() && !ShuntingYard::isNumber(output.back(), true, false))
            output.erase(output.size() - 1, 1);
    }
    else
    {
        while(!output.empty() && !ShuntingYard::isNumber(output.back(), false, false))
            output.erase(output.size() - 1, 1);
    }


    // Is the result still a number? It might not be if the input was an
    // enumeration or defined constant or some such. In that case we just
    // apply a cast and hope for the best.
    bool ok;
    double value = ShuntingYard::toNumber(output, &ok);
    if(!ok)
        return "(" + toTypeString() + ")" + number;

    // Add the correct suffix for the numeric type
    if(isFloat)
    {
        // Make sure we have a decimal point. If we already have one, but its
        // the last character, add the 0. So "10" -> "10.0", or "10." -> "10.0"
        if(output.find('.') == std::string::npos)
            output += ".0";
        else if(output.back() == '.')
            output += "0";

        // Finally put the 'f' at the end if this is a float
        if(bits <= 32)
            output += "f";
    }
    else
    {
        std::string suffix;

        // In this case we only need the suffix if the integer constant is large enough
        value = fabs(value);

        if(value >= 4294967295.0)
            suffix = "ll";
        else if((value >= 2147483647.0) && isSigned)
            suffix = "ll";
        else if(value >= 65535.0)
            suffix = "l";
        else if((value >= 32767.0) && isSigned)
            suffix = "l";

        if(!suffix.empty())
        {
            if(!isSigned)
                output += "u";

            output += suffix;
        }
    }

    return output;

}// TypeData::applyTypeToConstant


/*!
 * Pull a positive integer value from a string
 * \param string is the source string, which can contain a decimal or hexadecimal (0x) value
 * \param ok receives false if there was a problem with the decode
 * \return the positive integer value or 0 if there was a problem
 */
int TypeData::extractPositiveInt(const std::string& string, bool* ok)
{
    int base = 10;
    std::regex regex;
    int value = 0;

    if(contains(string, "0b"))
    {
        base = 2;
        regex = std::regex(R"([^01])");
    }
    else if(contains(string, "0x"))
    {
        base = 16;
        regex = std::regex(R"([^0123456789AaBbCcDdEeFf])");
    }
    else
    {
        base = 10;
        regex = std::regex(R"([^0123456789])");
    }

    if(ok != nullptr)
        (*ok) = true;

    try
    {
        value = std::stoi(std::regex_replace(string, regex, ""), nullptr, base);
    }
    catch (...)
    {
        value = 0;
        if(ok != nullptr)
            (*ok) = false;
    }

    return value;
}


/*!
 * Pull a double precision value from a string
 * \param string is the source string
 * \param ok receives false if there was a problem with the decode
 * \return the double precision value or 0 if there was a problem
 */
double TypeData::extractDouble(const std::string& string, bool* ok)
{
    int base = 10;
    std::regex regex;
    double value = 0;

    if(contains(string, "0b") && !contains(string, "."))
    {
        base = 2;
        regex = std::regex(R"([^01])");
    }
    else if(contains(string, "0x") && !contains(string, "."))
    {
        base = 16;
        regex = std::regex(R"([^0123456789AaBbCcDdEeFf])");
    }
    else
    {
        base = 10;
        regex = std::regex(R"([^0123456789\-\.])");
    }

    if(ok != nullptr)
        (*ok) = true;

    try
    {
        if(base == 10)
            value = std::stod(std::regex_replace(string, regex, ""), nullptr);
        else
            value = std::stol(std::regex_replace(string, regex, ""), nullptr, base);
    }
    catch (...)
    {
        value = 0;
        if(ok != nullptr)
            (*ok) = false;
    }

    return value;
}


/*!
 * Construct a blank protocol field
 * \param parse points to the global protocol parser that owns everything
 * \param parent is the hierarchical name of the parent object
 * \param supported indicates what the protocol can support
 */
ProtocolField::ProtocolField(ProtocolParser* parse, std::string parent, ProtocolSupport supported):
    Encodable(parse, parent, supported),
    encodedMin(0),
    encodedMax(0),
    scaler(1),
    hasVerifyMinValue(false),
    verifyMinValue(0),
    limitMinValue(0),
    hasVerifyMaxValue(false),
    verifyMaxValue(0),
    limitMaxValue(0),
    checkConstant(false),
    overridesPrevious(false),
    isOverriden(false),
    inMemoryType(supported),
    encodedType(supported),
    prevField(0),
    hidden(false),
    neverOmit(false),
    mapOptions(MAP_BOTH)
{
    attriblist = {"name",
                  "title",
                  "inMemoryType",
                  "encodedType",
                  "struct",
                  "max",
                  "min",
                  "scaler",
                  "printscaler",
                  "array",
                  "variableArray",
                  "array2d",
                  "variable2dArray",
                  "dependsOn",
                  "dependsOnValue",
                  "dependsOnCompare",
                  "enum",
                  "default",
                  "constant",
                  "checkConstant",
                  "comment",
                  "Units",
                  "Range",
                  "Notes",
                  "bitfieldGroup",
                  "hidden",
                  "neverOmit",
                  "initialValue",
                  "verifyMinValue",
                  "verifyMaxValue",
                  "map",
                  "limitOnEncode"};
}


/*!
 * Reset all data to defaults
 */
void ProtocolField::clear(void)
{
    Encodable::clear();

    encodedMin = encodedMax = 0;
    scaler = 1;
    defaultString.clear();
    defaultStringForDisplay.clear();
    constantString.clear();
    constantStringForDisplay.clear();
    checkConstant = false;
    overridesPrevious = false;
    isOverriden = false;
    encodedType = inMemoryType = TypeData(support);
    bitfieldData.clear();
    scalerString.clear();
    printScalerString.clear();
    readScalerString.clear();
    minString.clear();
    maxString.clear();
    prevField = 0;
    extraInfoNames.clear();
    extraInfoValues.clear();
    hidden = false;
    neverOmit = false;
    initialValueString.clear();
    verifyMinString.clear();
    verifyMaxString.clear();
    initialValueStringForDisplay.clear();
    verifyMinStringForDisplay.clear();
    verifyMaxStringForDisplay.clear();
    limitMinString.clear();
    limitMinStringForComment.clear();
    limitMinValue = 0;
    hasVerifyMinValue = false;
    verifyMinValue = 0;
    limitMaxString.clear();
    limitMaxStringForComment.clear();
    limitMaxValue = 0;
    hasVerifyMaxValue = false;
    verifyMaxValue = 0;

}// ProtocolField::clear


//! Provide the pointer to a previous encodable in the list
void ProtocolField::setPreviousEncodable(Encodable* prev)
{
    prevField = NULL;

    // We need to know when the bitfields end
    if(prev != NULL)
    {
        // Does prev point to a ProtocolField?
        prevField = dynamic_cast<ProtocolField*>(prev);
    }

    if(prevField == NULL)
        return;

    // Are we the start of part of a new bitfield group (or are we not a bitfield at all)?
    // Which means the previous field terminates that group (if any)
    if(bitfieldData.groupStart || !encodedType.isBitfield)
        prevField->setTerminatesBitfield(true);

    if(prevField->isBitfield() && (encodedType.isBitfield))
    {
        // Are we part of a bitfield group?
        if(!bitfieldData.groupStart)
        {
            // We did not start a group, we might be a member of a previous group
            bitfieldData.groupMember = prevField->bitfieldData.groupMember;

            // Previous bitfield does not terminate the bitfields
            prevField->setTerminatesBitfield(false);

            // We start at some nonzero bitcount that continues from the previous
            setStartingBitCount(prevField->getEndingBitCount());

            // Now that we know our starting bitcount we can apply the bitfield defaults warning
            if(!defaultString.empty() && (defaultString != "0") && (bitfieldData.startingBitCount != 0) && !bitfieldCrossesByteBoundary())
            {
                // The key concept is this: bitfields can have defaults, but if
                // the bitfield does not start a new byte it is possible (not
                // guaranteed) that the default will be overwritten because the
                // previous bitfield will have caused the packet length to
                // satisfy the length requirement. As an example:
                //
                // Old packet = 1 byte + 1 bit.
                // New augmented packet = 1 byte + 2 bits (second bit has a default).
                //
                // The length of the old and new packets is the same (2 bytes)
                // which means the default value for the augmented bit will
                // always be overwritten with zero.
                emitWarning("Bitfield with non-zero default which does not start a new byte; default may not be obeyed");
            }

        }

    }// if previous and us are bitfields

    computeEncodedLength();

}// ProtocolField::setPreviousEncodable


/*!
 * Get overriden type information.
 * \param prev is the previous encodable to test if its the source of the data being overriden by this encodable. Can be null
 * \return true if prev is the source of data being overriden
 */
bool ProtocolField::getOverriddenTypeData(ProtocolField* prev)
{
    // If we are overriding then this function is not interesting
    if(!overridesPrevious)
        return false;

    // Check to make sure that this previous actually exists
    if(prev == NULL)
        return false;

    // Must have the same name if we are overriding it
    if(prev->name != name)
        return false;

    // Must exist in memory, or we can't be overriding it
    if(prev->isNotInMemory())
        return false;

    // Let the previous one know that we are overriding it
    prev->isOverriden = true;

    // If we get here, then this is our baby. Update the data being overriden.
    inMemoryType = prev->inMemoryType;

    if(!inMemoryType.enumName.empty())
        emitWarning("Enumeration name ignored for overridden field");

    inMemoryType.enumName = prev->inMemoryType.enumName;

    if(!array.empty())
        emitWarning("Array information ignored for overridden field");
    array = prev->array;

    if(!array2d.empty())
        emitWarning("2D Array information ignored for overridden field");
    array2d = prev->array2d;

    // This information can be modified, but is typically taken from the original.
    if(variableArray.empty())
        variableArray = prev->variableArray;

    if(variable2dArray.empty())
        variable2dArray = prev->variable2dArray;

    if(dependsOn.empty())
        dependsOn = prev->dependsOn;

    if(comment.empty())
        comment = prev->comment;

    // Recompute the length now that the array data are up to date
    computeEncodedLength();

    return true;

}// ProtocolField::getOverriddenTypeData


//! Get the maximum number of temporary bytes needed for a bitfield group of our children
void ProtocolField::getBitfieldGroupNumBytes(int* num) const
{
    if(encodedType.isBitfield && bitfieldData.lastBitfield && bitfieldData.groupMember)
    {
        int length = ((bitfieldData.groupBits+7)/8);

        if(length > (*num))
            (*num) = length;
    }

}


/*!
 * Extract the type information from the type string, for in memory types
 * \param data holds the extracted type
 * \param type is the type string
 * \param inMemory is true if this is an in-memory type string, else encoded
 * \param _enumName is the name of the enumeration, if this is an enumerated type.
 */
void ProtocolField::extractType(TypeData& data, const std::string& typeString, bool inMemory, const std::string& _enumName)
{
    std::string type(typeString);

    data.clear();

    if(startsWith(type, "n"))
        data.isNull = true;
    else if(startsWith(type, "over") && inMemory)
    {
        overridesPrevious = true;

        // This is just a place holder, it will get overriden later
        data.bits = 32;
    }
    else if(startsWith(type, "stru"))
    {
        if(inMemory)
            data.isStruct = true;
        else
            return;
    }
    else if(startsWith(type, "string"))
    {
        data.isString = true;
        data.isFixedString = false;

        data.bits = 8;
    }
    else if(startsWith(type, "fixedstring"))
    {
        data.isString = true;
        data.isFixedString = true;

        data.bits = 8;
    }
    else if(startsWith(type, "bo"))
    {
        // default to unsigned 8
        data.bits = 8;

        if(support.supportbool == false)
        {
            emitWarning("bool support is disabled in this protocol");
        }
        else if(!inMemory)
        {
            emitWarning("bool is not a valid encoded type - it can only be used for in-memory types");
        }
        else
        {
            data.isBool = true;
        }
    }
    else if(startsWith(type, "bi"))
    {
        // Get the number of bits, between 1 and 32 inclusive
        data.bits = data.extractPositiveInt(type);

        if(support.bitfield == false)
        {
            emitWarning("bitfield support is disabled in this protocol");

            // if bits is 1, then it becomes 8. If it is 8 then it
            // becomes 8, if its 9 it becomes 16, etc.
            data.bits = 8*((data.bits + 7)/8);
        }
        else
        {
            // Bitfields cannot be floats and cannot be signed
            data.isBitfield = true;

            // Bitfields must have at least one bit, and less than 33 bits
            if(data.bits < 1)
            {
                data.bits = 1;
                emitWarning("bitfields must have a bit width of at least one");
            }
            else if((data.bits > 32) && (support.longbitfield == false))
            {
                emitWarning("bitfields must have a bit width of 32 or less");
                data.bits = 32;
            }
            else if(data.bits > 64)
            {
                emitWarning("bitfields must have a bit width of 64 or less");
                data.bits = 64;
            }
        }
    }
    else if(startsWith(type, "e"))
    {
        // enumeration types are only for in-memory, never encoded
        if(inMemory)
        {
            data.isEnum = true;
            data.enumName = _enumName;
        }
        else
        {
            data.isEnum = false;
        }

        data.bits = 8;
    }
    else
    {
        // Get the number of bits, between 1 and 32 inclusive
        data.bits = data.extractPositiveInt(type);

        if(startsWith(type, "u"))
        {
            data.isSigned = false;
        }
        else
        {
            data.isSigned = true;

            if(startsWith(type, "f"))
            {
                // we want to handle the cas where the user types "float16:10" to specify the number of significands
                if(type.find(':') != std::string::npos)
                {
                    std::vector<std::string> list = split(type, ":");

                    if(list.size() >= 2)
                    {
                        data.bits = data.extractPositiveInt(list.at(0));
                        data.sigbits = data.extractPositiveInt(list.at(1));
                    }

                }

                data.isFloat = true;

                // "float" is not a warning
                if(data.bits == 0)
                    data.bits = 32;
            }
            else if(startsWith(type, "d"))
            {
                data.isFloat = true;

                // "double" is not a warning
                if(data.bits == 0)
                    data.bits = 64;
            }
            else if(!startsWith(type, "s") && !startsWith(type, "i"))
            {
                emitWarning("in memory type name not understood, signed integer assumed");
            }
        }

        if(data.isFloat)
        {
            if(inMemory)
            {
                if((data.bits != 32) && (data.bits != 64))
                {
                    emitWarning("in memory float types must be 32 or 64 bits");

                    if(data.bits < 32)
                        data.bits = 32;
                    else
                        data.bits = 64;
                }

                if(data.sigbits != 0)
                {
                    data.sigbits = 0;
                    emitWarning("in memory float types do not have variable resolution");
                }
            }
            else
            {
                if((data.bits != 16) && (data.bits != 24) && (data.bits != 32) && (data.bits != 64))
                {
                    emitWarning("encoded float types must be 16, 24, 32, or 64 bits");

                    if(data.bits < 16)
                        data.bits = 16;
                    else if(data.bits < 24)
                        data.bits = 24;
                    else if(data.bits < 32)
                        data.bits = 32;
                    else
                        data.bits = 64;

                    if((data.bits < 32) && (support.specialFloat == false))
                    {
                        emitWarning("non-standard float bit widths are disabled in this protocol");
                        data.bits = 32;
                        data.sigbits = 0;
                    }
                }

                if(data.sigbits != 0)
                {
                    if(data.bits >= 32)
                    {
                        emitWarning("float type must be 16 or 24 bit to specify resolution");
                        data.sigbits = 0;
                    }
                    else if(data.bits == 24)
                    {
                        if((data.sigbits < 4) || (data.sigbits > 20))
                        {
                            emitWarning("significand (resolution) of float24 must be between 4 and 20 bits inclusive, defaulted to 15");
                            data.sigbits = 15;
                        }
                    }
                    else if(data.bits == 16)
                    {
                        if((data.sigbits < 4) || (data.sigbits > 12))
                        {
                            emitWarning("significand (resolution) of float16 must be between 4 and 12 bits inclusive, defaulted to 9");
                            data.sigbits = 9;
                        }
                    }

                }// if user specified significant bits
                else
                {
                    // Default significand bits for float16 and float24
                    if(data.bits == 16)
                        data.sigbits = 9;
                    else if(data.bits == 24)
                        data.sigbits = 15;
                }

            }// else if not in memory float

            if((data.bits > 32) && (support.float64 == false))
            {
                emitWarning("64 bit float support is disabled in this protocol");
                data.bits = 32;
            }

        }// if float
        else
        {
            if(inMemory)
            {
                if((data.bits != 8) && (data.bits != 16) && (data.bits != 32) && (data.bits != 64))
                {
                    emitWarning("in memory integer types must be 8, 16, 32, or 64 bits");

                    if(data.bits > 32)
                        data.bits = 64;
                    else if(data.bits > 16)
                        data.bits = 32;
                    else if(data.bits > 8)
                        data.bits = 16;
                    else
                        data.bits = 8;
                }
            }
            else if(((data.bits % 8) != 0) || (data.bits > 64))
            {
                emitWarning("encoded integer types must be 8, 16, 24, 32, 40, 48, 56, or 64 bits");

                if(data.bits > 56)
                    data.bits = 64;
                else if(data.bits > 48)
                    data.bits = 56;
                else if(data.bits > 40)
                    data.bits = 48;
                else if(data.bits > 32)
                    data.bits = 40;
                else if(data.bits > 24)
                    data.bits = 32;
                else if(data.bits > 16)
                    data.bits = 24;
                else if(data.bits > 8)
                    data.bits = 16;
                else
                    data.bits = 8;
            }

            if((data.bits > 32) && (support.int64 == false))
            {
                emitWarning("Integers greater than 32 bits are disabled in this protocol");
                data.bits = 32;
            }

        }// if integer

    }// else if float or integer

}// ProtocolField::extractType


/*!
 * Parse the DOM to determine the details of this ProtocolField
 */
void ProtocolField::parse(void)
{
    std::string memoryTypeString;
    std::string encodedTypeString;
    std::string structName;
    std::string enumName;

    clear();

    if(e == nullptr)
        return;

    const XMLAttribute* map = e->FirstAttribute();

    // We use name as part of our debug outputs, so its good to have it first.
    name = ProtocolParser::getAttribute("name", map);

    // Tell the user of attribute problems
    testAndWarnAttributes(map);

    // This will propagate to any of the children we create
    if(ProtocolParser::isFieldSet("limitOnEncode", map))
        support.limitonencode = true;
    else if(ProtocolParser::isFieldClear("limitOnEncode", map))
        support.limitonencode = false;

    title = ProtocolParser::getAttribute("title", map);
    memoryTypeString = ProtocolParser::getAttribute("inMemoryType", map);
    encodedTypeString = ProtocolParser::getAttribute("encodedType", map);
    structName = ProtocolParser::getAttribute("struct", map);
    maxString = ProtocolParser::getAttribute("max", map);
    minString = ProtocolParser::getAttribute("min", map);
    scalerString = ProtocolParser::getAttribute("scaler", map);
    printScalerString = ProtocolParser::getAttribute("printscaler", map);
    array = ProtocolParser::getAttribute("array", map);
    variableArray = ProtocolParser::getAttribute("variableArray", map);
    array2d = ProtocolParser::getAttribute("array2d", map);
    variable2dArray = ProtocolParser::getAttribute("variable2dArray", map);
    dependsOn = ProtocolParser::getAttribute("dependsOn", map);
    dependsOnValue = ProtocolParser::getAttribute("dependsOnValue", map);
    dependsOnCompare = ProtocolParser::getAttribute("dependsOnCompare", map);
    enumName = ProtocolParser::getAttribute("enum", map);
    defaultString = ProtocolParser::getAttribute("default", map);
    constantString = ProtocolParser::getAttribute("constant", map);
    comment = ProtocolParser::getAttribute("comment", map);
    bitfieldData.groupMember = bitfieldData.groupStart = ProtocolParser::isFieldSet("bitfieldGroup", map);
    initialValueString = ProtocolParser::getAttribute("initialValue", map);
    verifyMinString = ProtocolParser::getAttribute("verifyMinValue", map);
    verifyMaxString = ProtocolParser::getAttribute("verifyMaxValue", map);
    hidden = ProtocolParser::isFieldSet("hidden", map);
    neverOmit = ProtocolParser::isFieldSet("neverOmit", map);
    checkConstant = ProtocolParser::isFieldSet("checkConstant", map);

    extraInfoValues.push_back(ProtocolParser::getAttribute("Units", map));
    if(extraInfoValues.back().empty())
        extraInfoValues.pop_back();
    else
        extraInfoNames.push_back("Units");

    extraInfoValues.push_back(ProtocolParser::getAttribute("Range", map));
    if(extraInfoValues.back().empty())
        extraInfoValues.pop_back();
    else
        extraInfoNames.push_back("Range");

    extraInfoValues.push_back(ProtocolParser::getAttribute("Notes", map));
    if(extraInfoValues.back().empty())
        extraInfoValues.pop_back();
    else
        extraInfoNames.push_back("Notes");

    std::string temp = ProtocolParser::getAttribute("map", map);
    if(!temp.empty())
    {
        if(contains(temp, "encode"))
            mapOptions = MAP_ENCODE;
        else if(contains(temp, "decode"))
            mapOptions = MAP_DECODE;
        else if(ProtocolParser::isFieldSet(temp))
            mapOptions = MAP_BOTH;
        else if(ProtocolParser::isFieldClear(temp))
            mapOptions = MAP_NONE;
        else
            emitWarning("Value for 'map' field is incorrect: '" + temp + "'");

    }

    if(name.empty() && (memoryTypeString != "null"))
    {
        emitWarning("Data tag without a name");
    }

    if(title.empty())
        title = name;

    // maybe its an enum or a external struct?
    if(memoryTypeString.empty())
    {
        // Maybe its an enum?
        if(e->Attribute("enum"))
            memoryTypeString = "enum";
        else if(e->Attribute("struct"))
            memoryTypeString = "struct";
        else
        {
            memoryTypeString = "null";
            emitWarning("failed to find inMemoryType attribute, \"null\" assumed.");
        }
    }

    // Extract the in memory type
    extractType(inMemoryType, memoryTypeString, true, enumName);

    // The encoded type string, this can be empty which implies encoded is same as memory
    if(encodedTypeString.empty())
    {
        if(overridesPrevious)
            emitWarning("encodedType cannot be empty if inMemoryType is override");

        encodedType = inMemoryType;

        // Encoded types are never enums
        if(encodedType.isEnum)
            encodedType.isEnum = false;
    }
    else
    {
        extractType(encodedType, encodedTypeString, false);

        // This is just a warning pacifier, we won't learn until later what the
        // in memory type is
        if(overridesPrevious)
            inMemoryType = encodedType;
    }

    // If we have a hidden field, and we are not supposed to omit code for that
    // field, we treat it as null. This is because we *must* omit the code in
    // order to adhere to the encoding packet rules. This may break some code
    // if other fields depend on the hidden field.
    if(hidden && !neverOmit && support.omitIfHidden)
    {
        comment = "Hidden field skipped";
        inMemoryType.clear();
        inMemoryType.isNull = true;

        // No scaling
        scalerString.clear();
        maxString.clear();
        minString.clear();
        verifyMaxString.clear();
        verifyMinString.clear();
        printScalerString.clear();

        // This is not a warning, just useful information
        std::cout << "Skipping code output for hidden field " << getHierarchicalName() << std::endl;
    }

    if(inMemoryType.isNull)
    {
        // Null types are not in memory, therefore cannot have defaults or variable arrays
        /// TODO: is this strictly true? It seems we could relax this requirement somewhat
        variableArray.clear();
        variable2dArray.clear();
        defaultString.clear();
        overridesPrevious = false;

        // A special case, where we use the encoded type data in place of the
        // in memory type. This handles cases where (for example) we want to
        // encode a constant bitfield, or a string, but have no data in memory
        if(!encodedType.isNull)
        {
            inMemoryType = encodedType;
            inMemoryType.isNull = true;
        }
        else
        {
            emitWarning("both in-memory and encoded types are \"null\", nothing to do.");
            return;
        }
    }

    if(inMemoryType.isEnum)
    {
        if(inMemoryType.enumName.empty())
        {
            emitWarning("enumeration name is missing, type changed to unsigned");
            inMemoryType.isEnum = encodedType.isEnum = false;
        }
        else
        {
            int minbits = 8;
            inMemoryType.enummax = 255;

            // Figure out the minimum number of bits for the enumeration
            const EnumCreator* creator = parser->lookUpEnumeration(inMemoryType.enumName);
            if(creator != nullptr)
            {
                minbits = creator->getMinBitWidth();
                inMemoryType.enummax = creator->getMaximumValue();
            }

            if(encodedTypeString.empty())
            {
                // Make it a multiple of 8 bits. The only way to have something
                // different is to encode as a bitfield, which means the
                // encoded string won't be empty
                if(minbits % 8)
                    minbits = ((minbits/8)+1)*8;

                encodedType.bits = minbits;

            }// if we don't have encoded length data
            else
            {
                // make sure the encoded length data is large enough
                if(encodedType.bits < minbits)
                {
                    emitWarning("enumeration needs at least " + std::to_string(minbits) + " bits. Encoded bit length changed.");
                    if(!encodedType.isBitfield)
                    {
                        // Make it a multiple of 8 bits
                        if(minbits % 8)
                            minbits = ((minbits/8)+1)*8;
                    }

                    encodedType.bits = minbits;

                }// if not enough encoded bits

            }// else we do have encoded length data

        }// if we have the enumeration name

    }// if in memory data is enumeration

    if(inMemoryType.isStruct)
    {
        if(structName.empty())
        {
            emitWarning("struct name is missing, struct name \"unknown\" used, probable compile failure");
            structName = "unknown";
        }

        if(!constantString.empty() || checkConstant)
        {
            constantString.clear();
            checkConstant = false;
            emitWarning("structure cannot be a constant");
        }

        if(overridesPrevious)
        {
            overridesPrevious = false;
            emitWarning("structure cannot override a previous field");
        }

    }

    if(inMemoryType.isBitfield)
    {
        if(!encodedType.isNull)
        {
            if(!encodedTypeString.empty() && !encodedType.isBitfield)
                emitWarning("encoded type ignored because in memory type is bitfield");

            // make the encoded type follow the in memory type for bit fields
            encodedType.isBitfield = true;
            encodedType.bits = inMemoryType.bits;
        }
    }

    // It is possible for the in memory type to not be a bit field, but the
    // encoding could be. The most common case for this would be an in-memory
    // enumeration in which the maximum enumeration fits in fewer than 8 bits
    if(encodedType.isBitfield)
    {
        // Do we start a bitfield group?
        if(bitfieldData.groupMember)
        {
            if(!defaultString.empty())
            {
                emitWarning("bitfield groups cannot have default values");
                defaultString.clear();
            }

        }

        if(!dependsOn.empty())
        {
            emitWarning("bitfields cannot use dependsOn");
            dependsOn.clear();
        }

        if(!array.empty())
        {
            emitWarning("bitfields encodings cannot use arrays");
            array.clear();
            variableArray.clear();
            array2d.clear();
            variable2dArray.clear();
        }

        // We assume we are the last member of the bitfield, until we learn otherwise
        bitfieldData.lastBitfield = true;
    }
    else
    {
        if(bitfieldData.groupMember || bitfieldData.groupStart)
        {
            emitWarning("bitfieldGroup applied to non-bitfield, ignored");
            bitfieldData.groupBits = 0;
            bitfieldData.groupStart = false;
            bitfieldData.groupMember = false;
        }
    }

    // if either type says string, than they both are string
    if(inMemoryType.isString != encodedType.isString)
    {
        if(!inMemoryType.isNull && !encodedType.isNull)
            emitWarning("String type requires that inMemory and encoded types both be strings");

        inMemoryType.isString = encodedType.isString = true;
        inMemoryType.bits = encodedType.bits = 8;
    }

    // if either type says fixed string, than they both are fixed string
    if(inMemoryType.isFixedString != encodedType.isFixedString)
    {
        inMemoryType.isString = encodedType.isString = true;
        inMemoryType.isFixedString = encodedType.isFixedString = true;
        inMemoryType.bits = encodedType.bits = 8;
    }

    if(array.empty() && !variableArray.empty())
    {
        emitWarning("Must specify array length to specify variable array length");
        variableArray.clear();
    }

    if(array.empty() && !array2d.empty())
    {
        emitWarning("Must specify array length to specify second dimension array length");
        array2d.clear();
    }

    if(array2d.empty() && !variable2dArray.empty())
    {
        emitWarning("Must specify array 2d length to specify variable 2d array length");
        variable2dArray.clear();
    }

    if(!scalerString.empty() && !maxString.empty())
    {
        emitWarning("scaler ignored because max is provided");
        scalerString.clear();
    }

    // handle the scaler used for printing and comparing
    if(!printScalerString.empty())
    {
        bool ok;
        double printScaler = ShuntingYard::computeInfix(printScalerString, &ok);

        if(ok)
        {
            printScalerString = "*" + getNumberString(printScaler);
            readScalerString = "/" + getNumberString(printScaler);
        }
        else
        {
            emitWarning("Print scaler is not a number, scaling during printing or comparing will not be done");
            printScalerString = "";
            readScalerString = "";
        }

    }

    if(inMemoryType.isBool)
    {
        if(encodedType.isBitfield && (encodedType.bits > 1))
            emitWarning("Boolean in memory is encoded to bitfield larger than 1 bit");
        else if(encodedType.isFloat)
        {
            emitWarning("Boolean cannot be encoded as float");
            encodedType.isFloat = 0;
        }
    }

    if(!maxString.empty() || !minString.empty() || !scalerString.empty())
    {
        if(inMemoryType.isStruct || inMemoryType.isString || encodedType.isNull || inMemoryType.isBool)
        {
            emitWarning("min, max, and scaler do not apply to this type data");
            maxString.clear();
            minString.clear();
            scalerString.clear();
        }
    }

    if(!maxString.empty() || !minString.empty())
    {
        if(encodedType.isFloat)
        {
            emitWarning("min, max, are ignored because encoded type is float");
            maxString.clear();
            minString.clear();
        }
    }

    if(!minString.empty())
    {
        if(encodedType.isSigned)
        {
            emitWarning("min value ignored because encoded type is signed");
            minString.clear();
        }
    }

    if(constantString.empty() && checkConstant)
    {
        emitWarning("\"checkConstant\" cannot be applied unless the field is constant");
        checkConstant = false;
    }

    if(inMemoryType.isString)
    {
        // Strings have to be arrays, default to 64 characters
        if(array.empty())
        {
            emitWarning("string length not provided, assuming 64");
            array = "64";
        }

        // Strings are always variable length, through null termination
        if(!variableArray.empty())
        {
            emitWarning("strings cannot use variableAray attribute, they are always variable length through null termination (unless fixedstring)");
            variableArray.clear();
        }

        if(!array2d.empty())
        {
            emitWarning("2d arrays not allowed for strings");
            array2d.clear();
            variable2dArray.clear();
        }

        if(!dependsOn.empty())
        {
            emitWarning("strings cannot use dependsOn");
            dependsOn.clear();
        }

    }// if string
    else if(!array.empty())
    {
        if(checkConstant)
        {
            emitWarning("\"checkConstant\" cannot be applied to arrays (except strings) ");
            checkConstant = false;
        }

    }// if not string

    if(encodedType.isNull)
    {
        if(!constantString.empty())
        {
            emitWarning("constant value does not make sense for types that are not encoded (null)");
            constantString.clear();
            checkConstant = false;
        }

        if(!variableArray.empty() || !variable2dArray.empty())
        {
            emitWarning("variable length arrays do not make sense for types that are not encoded (null)");
            variableArray.clear();
            variable2dArray.clear();
        }

        if(!dependsOn.empty())
        {
            emitWarning("dependsOn does not make sense for types that are not encoded (null)");
            dependsOn.clear();
        }
    }

    if(!dependsOnValue.empty() && dependsOn.empty())
    {
        emitWarning("dependsOnValue does not make sense unless dependsOn is defined");
        dependsOnValue.clear();
    }

    if(!dependsOnCompare.empty() && dependsOnValue.empty())
    {
        emitWarning("dependsOnCompare does not make sense unless dependsOnValue is defined");
        dependsOnCompare.clear();
    }
    else if(dependsOnCompare.empty() && !dependsOnValue.empty())
    {
        // This is not a warning, it is expected
        dependsOnCompare = "==";
    }

    bool ok;

    // minString specifies the minimum value that can be encoded. It only
    // applies if the encoded type is not signed, and if the encoded type is
    // not floating point. It is possible to have a minString without a
    // scaler or maxString.
    if(!minString.empty())
    {
        encodedMin = ShuntingYard::computeInfix(parser->replaceEnumerationNameWithValue(minString), &ok);

        if(!ok)
        {
            emitWarning("min is not a number, 0.0 assumed");
            encodedMin = 0.0;
            minString = "0";
        }
        else
        {
            // Special case handling here, if the user just wants a min value, that implies that scaler is 1.0
            if(maxString.empty() && scalerString.empty())
                scalerString = "1";
        }
    }

    // maxString specifies the maximum value that can be encoded. It only
    // applies if the encoded type is not floating point. It is possible
    // to have a maxString without a scaler or minString. However if no scaler
    // is specified then one will be computed based on maxString.
    if(!maxString.empty())
    {
        encodedMax = ShuntingYard::computeInfix(parser->replaceEnumerationNameWithValue(maxString), &ok);
        if(!ok)
        {
            emitWarning("max is not a number, 1.0 assumed");
            encodedMax = 1.0;
            maxString = "1";
        }

        // We know at this point that the encodedType is not floating point
        if(encodedType.isSigned)
        {
            // Reconstruct a scaler based on the encodedMax
            scaler = (encodedType.getMaximumIntegerValue())/encodedMax;
            scalerString = std::to_string(encodedType.getMaximumIntegerValue()) + "/(" + maxString + ")";

            // This is not exactly true, there is one more bit that could be used,
            // but this makes conciser commenting, and is clearer to the user
            encodedMin = -encodedMax;
            minString = "-" + maxString;

        }// if encoded type is signed
        else
        {
            // Reconstruct a scaler based on the encodedMax and encodedMin
            scaler = (encodedType.getMaximumIntegerValue())/(encodedMax - encodedMin);

            // The string we use for documentation should be made as reasonably concise as possible
            if(encodedMin == 0.0)
            {
                minString = "0";
                scalerString = std::to_string(encodedType.getMaximumIntegerValue())+ "/(" + maxString + ")";
            }
            else
            {
                // If the user gives us something like 145 for max and -5 for min, I would rather just put 150 in the documentation
                std::string denominator = "(" + maxString + " - " + minString + ")";

                // Documentation is only improved if maxString and minString are simple numbers, not formulas
                if(ShuntingYard::isNumber(maxString) && ShuntingYard::isNumber(minString))
                {
                    bool ok;
                    double number = ShuntingYard::computeInfix(denominator, &ok);
                    if(ok)
                        denominator = getNumberString(number);
                }

                scalerString = std::to_string(encodedType.getMaximumIntegerValue())+ "/" + denominator;
            }

        }// else if encoded type is unsigned

    }// if a maxString was specified
    else if(!scalerString.empty())
    {
        // scaler specifies the multiplier to apply to the in-Memory value to
        // produce the encoded value. Unlike minString and maxString it can
        // apply even if the encodedType is floating point. Hence it is
        // possible for encodedMax to be 0 while the scaler is not 1.0.
        scaler = ShuntingYard::computeInfix(parser->replaceEnumerationNameWithValue(scalerString), &ok);

        if(!ok)
        {
            emitWarning("scaler is not a number, 1.0 assumed");
            scaler = 1.0;
            scalerString = "1.0";
        }
        else if(scaler <= 0.0)
        {
            emitWarning("scaler must be greater than zero, 1.0 used");
            scaler = 1.0;
            scalerString = "1.0";
        }

        if(encodedType.isFloat)
        {
            // Floating point scaling does not have min and max.
            encodedMax = encodedMin = 0;
            minString = maxString = "";
        }
        else if(encodedType.isSigned)
        {
            encodedMax = (encodedType.getMaximumIntegerValue())/scaler;
            maxString = std::to_string(encodedType.getMaximumIntegerValue())+ "/(" + scalerString + ")";

            // This is not exactly true, there is one more bit that could be used,
            // but this makes conciser commenting, and is clearer to the user
            encodedMin = -encodedMax;
            minString = "-" + maxString;
        }
        else
        {
            encodedMax = encodedMin + (encodedType.getMaximumIntegerValue())/scaler;

            // Make sure minString isn't empty
            if(encodedMin == 0)
            {
                minString = "0";
                maxString = std::to_string(encodedType.getMaximumIntegerValue())+ "/(" + scalerString + ")";
            }
            else
                maxString = minString + " + " + std::to_string(encodedType.getMaximumIntegerValue())+ "/(" + scalerString + ")";
        }

    }// if scaler string

    // Max must be larger than minimum
    if(encodedMin > encodedMax)
    {
        encodedMin = encodedMax = 0.0;
        minString.clear();
        maxString.clear();
        scalerString.clear();
        scaler = 1.0;
        emitWarning("max is not more than min, encoding not scaled");
    }

    if(inMemoryType.isFloat && !encodedType.isFloat && !inMemoryType.isNull && !encodedType.isNull)
    {
        // If the user wants to convert a float to an integer they should be
        // applying a scaler. If they don't then there is the potential for
        // truncation and overflow problems. However its possible they actually
        // *want* the truncation, hence warning without fixing.
        if(encodedMin == encodedMax)
        {
            emitWarning("Casting float to integer without truncation, consider setting scaler=\"1.0\"");
        }
    }

    // Just the type data
    typeName = inMemoryType.toTypeString(support.prefix + structName);

    if(!constantString.empty())
    {
        if(inMemoryType.isStruct)
        {
            emitWarning("structure cannot have a constant value");
            constantString.clear();
        }
    }

    // Make sure no keyword conflicts
    checkAgainstKeywords();

    // Work out the limits that come from the encoding rules
    if(isFloatScaling() || isIntegerScaling())
    {
        limitMinValue = encodedMin;
        limitMaxValue = encodedMax;

        if(isFloatScaling())
        {
            limitMaxString = getNumberString(limitMaxValue, encodedType.bits);
            limitMinString = getNumberString(limitMinValue, encodedType.bits);
            limitMaxStringForComment = getNumberString(limitMaxValue);
            limitMinStringForComment = getNumberString(limitMinValue);
        }
        else
        {
            // Integer scaling is interesting. Imagine the case where we encode
            // in a signed16 number with a scaler of 8. Hence the largest value
            // after decode will be 32767/8 = 4095.875. However if the in-Memory
            // type is an integer then this will truncate to 4095, therefore use
            // of the trunc() function
            if(limitMaxValue >= 0)
                limitMaxStringForComment = limitMaxString = std::to_string((uint64_t)trunc(limitMaxValue));
            else
                limitMaxStringForComment = limitMaxString = std::to_string((int64_t)trunc(limitMaxValue));

            if(limitMinValue >= 0)
                limitMinStringForComment = limitMinString = std::to_string((uint64_t)trunc(limitMinValue));
            else
                limitMinStringForComment = limitMinString = std::to_string((int64_t)trunc(limitMinValue));
        }
    }
    else if(encodedType.isFloat)
    {
        limitMinValue = encodedType.getMinimumFloatValue()/scaler;
        limitMaxValue = encodedType.getMaximumFloatValue()/scaler;
        limitMaxString = getNumberString(limitMaxValue, encodedType.bits);
        limitMinString = getNumberString(limitMinValue, encodedType.bits);
        limitMaxStringForComment = getNumberString(limitMaxValue);
        limitMinStringForComment = getNumberString(limitMinValue);
    }
    else
    {
        limitMinValue = (double)encodedType.getMinimumIntegerValue();
        limitMaxValue = (double)encodedType.getMaximumIntegerValue();

        if(limitMaxValue >= 0)
            limitMaxStringForComment = limitMaxString = std::to_string((uint64_t)round(limitMaxValue));
        else
            limitMaxStringForComment = limitMaxString = std::to_string((int64_t)round(limitMaxValue));

        if(limitMinValue >= 0)
            limitMinStringForComment = limitMinString = std::to_string((uint64_t)round(limitMinValue));
        else
            limitMinStringForComment = limitMinString = std::to_string((int64_t)round(limitMinValue));
    }

    // Now handle the verify maximum value
    if(toLower(verifyMaxString) == "auto")
    {
        verifyMaxString = limitMaxString;
        hasVerifyMaxValue = true;
    }
    else if(!verifyMaxString.empty())
    {
        // Compute the verify max value if we can
        verifyMaxValue = ShuntingYard::computeInfix(parser->replaceEnumerationNameWithValue(verifyMaxString), &hasVerifyMaxValue);
    }

    // Now handle the verify minimum value
    if(toLower(verifyMinString) == "auto")
    {
        verifyMinString = limitMinString;
        hasVerifyMinValue = true;
    }
    else if(!verifyMinString.empty())
    {
        // Compute the verify min value if we can
        verifyMinValue = ShuntingYard::computeInfix(parser->replaceEnumerationNameWithValue(verifyMinString), &hasVerifyMinValue);
    }

    // Support the case where a numeric string uses "pi" or "e".
    defaultStringForDisplay = handleNumericConstants(defaultString);
    constantStringForDisplay = handleNumericConstants(constantString);
    verifyMaxStringForDisplay = handleNumericConstants(verifyMaxString);
    verifyMinStringForDisplay = handleNumericConstants(verifyMinString);
    initialValueStringForDisplay = handleNumericConstants(initialValueString);

    // Update these strings for display purposes
    maxString = handleNumericConstants(maxString);
    minString = handleNumericConstants(minString);
    scalerString = handleNumericConstants(scalerString);

    // Compute the data length
    computeEncodedLength();

}// ProtocolField::parse


/*!
 * Take an input string, which may be a number, and handle instances of "pi" or
 * "e", modifying the input string, and generating an output string which is
 * suitable for documentation purposes.
 * \param input is the input string, which may be modified
 * \return a documentation version of input
 */
std::string ProtocolField::handleNumericConstants(std::string& input) const
{
    if(input.empty())
        return input;

    bool ok;
    std::string display = input;

    // Determine if the input string is a number, which might have numeric constants (pi or e)
    ShuntingYard::computeInfix(input, &ok);

    if(ok)
    {
        // For the input string, we replace the symbols "pi" and "e" with their
        // numeric values so the code which uses this string will compile. We
        // cannot do this replacement without first knowing input is a number,
        // otherwise we'll just be screwing up the name of something
        ShuntingYard::replacePie(input);

        // For the display string we replace the symbol "pi" with the
        // appropriate value for html outputs
        replaceinplace(display, "pi", "&pi;");

        // Change to get rid of * multiply symbol, which plays havoc with markdown
        // We put spaces around the multiply symbol, so that the html tables can
        // better reflow the resulting text
        replaceinplace(display, "*", " &times; ");
    }

    return display;

}// ProtocolField::handleNumericConstants


/*!
 * Compute the encoded length of this field
 */
void ProtocolField::computeEncodedLength(void)
{
    encodedLength.clear();

    if(encodedType.isNull)
        return;

    if(encodedType.isBitfield)
    {
        if(bitfieldData.groupMember)
        {
            // If we are a group member, we need to figure out the number of bits in
            // the group, we can only do this if we are the last member of the group
            if(bitfieldData.lastBitfield)
            {
                int bits = 0;
                ProtocolField* prev = this;

                // Count backwards until our bitfield group ends in order to
                // determine the number of bits in the group
                while(prev != NULL)
                {
                    if(prev->encodedType.isBitfield && prev->bitfieldData.groupMember)
                    {
                        bits += prev->encodedType.bits;

                        if(prev->bitfieldData.groupStart)
                            break;
                        else
                            prev = prev->prevField;
                    }
                    else
                        break;
                }

                // Now we know the group bits, apply this to all members of the group
                prev = this;
                while(prev != NULL)
                {
                    if(prev->encodedType.isBitfield && prev->bitfieldData.groupMember)
                    {
                        prev->bitfieldData.groupBits = bits;

                        if(prev->bitfieldData.groupStart)
                            break;
                        else
                            prev = prev->prevField;
                    }
                    else
                        break;
                }

                // groupBits is visible to all fields in the group, but we only
                // want to count it once, so we only count for the lastBitfield
                encodedLength.addToLength(std::to_string((bitfieldData.groupBits+7)/8), false, false, false, !defaultString.empty());

            }// if we are the last member of the group

        }// if we are a group member
        else
        {
            int length = 0;

            // As a bitfield our length in bytes is given by the number of 8 bit boundaries we cross
            int bitcount = bitfieldData.startingBitCount;
            while(bitcount < getEndingBitCount())
            {
                bitcount++;
                if((bitcount % 8) == 0)
                    length++;
            }

            // If we are the last bitfield, and if we have any bits left, then add a byte
            if(bitfieldData.lastBitfield && ((bitcount % 8) != 0))
                length++;

            encodedLength.addToLength(std::to_string(length), false, false, false, !defaultString.empty());
        }
    }
    else if(inMemoryType.isString)
        encodedLength.addToLength(array, !inMemoryType.isFixedString, false, !dependsOn.empty(), !defaultString.empty());
    else if(inMemoryType.isStruct)
    {
        encodedLength.clear();

        const ProtocolStructure* struc = parser->lookUpStructure(typeName);

        // Account for array, variable array, and depends on
        if(struc != NULL)
            encodedLength.addToLength(struc->encodedLength, array, !variableArray.empty() || !variable2dArray.empty(), !dependsOn.empty(), array2d);
        else
        {
            if(is2dArray())
                encodedLength.addToLength("getMinLengthOf" + typeName + "()*" + array + "*" + array2d, false, !variableArray.empty() || !variable2dArray.empty(), !dependsOn.empty(), (!defaultString.empty()) || overridesPrevious);
            else if(isArray())
                encodedLength.addToLength("getMinLengthOf" + typeName + "()*" + array, false, !variableArray.empty(), !dependsOn.empty(), (!defaultString.empty()) || overridesPrevious);
            else
                encodedLength.addToLength("getMinLengthOf" + typeName + "()"         , false, false,                    !dependsOn.empty(), (!defaultString.empty()) || overridesPrevious);
        }
    }
    else
    {
        std::string lengthString = std::to_string(encodedType.bits / 8);

        // Remember that we could be encoding an array
        if(isArray())
            lengthString += "*" + array;

        if(is2dArray())
            lengthString += "*" + array2d;

        encodedLength.addToLength(lengthString, false, !variableArray.empty() || !variable2dArray.empty(), !dependsOn.empty(), (!defaultString.empty()) || overridesPrevious);

    }

}// ProtocolField::computeEncodedLength


//! Check names against the list of C keywords
void ProtocolField::checkAgainstKeywords(void)
{
    Encodable::checkAgainstKeywords();

    if(contains(keywords, inMemoryType.enumName, true))
    {
        emitWarning("enum name matches C keyword, changed to _name");
        inMemoryType.enumName = "_" + inMemoryType.enumName;
    }

    if(contains(keywords, maxString, true))
    {
        emitWarning("max value matches C keyword, changed to _max");
        maxString = "_" + maxString;
    }

    if(contains(keywords, minString, true))
    {
        emitWarning("min value matches C keyword, changed to _min");
        minString = "_" + minString;
    }

    if(contains(keywords, scalerString, true))
    {
        emitWarning("scaler value matches C keyword, changed to _scaler");
        scalerString = "_" + scalerString;
    }

    if(contains(keywords, defaultString, true))
    {
        emitWarning("default value matches C keyword, changed to _default");
        defaultString = "_" + defaultString;
    }

    if(contains(keywords, constantString, true))
    {
        emitWarning("constant value matches C keyword, changed to _constant");
        constantString = "_" + constantString;
    }

}// ProtocolField::checkAgainstKeywords


/*!
 * Get the declaration for this field as a member of a structure
 * \return the declaration string
 */
std::string ProtocolField::getDeclaration(void) const
{
    std::string output;

    if(isNotInMemory())
        return output;

    output = "    " + typeName + " " + name;

    if(inMemoryType.isBitfield)
        output += " : " + std::to_string(inMemoryType.bits);
    else if(is2dArray())
        output += "[" + array + "][" + array2d + "]";
    else if(isArray())
        output += "[" + array + "]";

    output += ";";

    if(comment.empty())
    {
        if(!constantString.empty())
            output += " //!< Field is encoded constant.";
    }
    else
    {
        output += " //!< " + comment;
        if(!constantString.empty())
            output += ". Field is encoded constant.";
    }

    output += "\n";

    return output;

}// ProtocolField::getDeclaration


/*!
 * push_back the include directives needed for this encodable. Mostly this is empty,
 * but for external structures or enumerations we need to bring in the include file
 * \param list is push_backed with any directives this encodable requires.
 */
void ProtocolField::getIncludeDirectives(std::vector<std::string>& list) const
{
    std::string include;

    // Array sizes could be enumerations that need an include directive
    if(!array.empty())
    {
        include = parser->lookUpIncludeName(array);
        if(!include.empty())
            list.push_back(include);

        // Array sizes could be enumerations that need an include directive
        if(!array2d.empty())
        {
            include = parser->lookUpIncludeName(array2d);
            if(!include.empty())
                list.push_back(include);
        }
    }

    if(inMemoryType.isEnum || inMemoryType.isStruct)
    {
        include = parser->lookUpIncludeName(typeName);
        if(include.empty())
        {
            // we don't warn for enumeration include failures, they are (potentially) less serious
            if(inMemoryType.isStruct)
                emitWarning("unknown include for \"" + typeName + "\"");
        }
        else
            list.push_back(include);

    }// if enum or struct

    // Only need one of each include
    removeDuplicates(list);

}// ProtocolField::getIncludeDirectives


/*!
 * push_back the include directives needed for this encodable. Mostly this is empty,
 * but for external structures or enumerations we need to bring in the include file
 * \param list is push_backed with any directives this encodable requires.
 */
void ProtocolField::getInitAndVerifyIncludeDirectives(std::vector<std::string>& list) const
{
    if(inMemoryType.isStruct)
    {
        const ProtocolStructureModule* struc = parser->lookUpStructure(typeName);

        if(struc != NULL)
            struc->getInitAndVerifyIncludeDirectives(list);

    }// if struct

}// ProtocolField::getInitAndVerifyIncludeDirectives


/*!
 * Return the include directives needed for this encodable's map functions
 * \param list is push_backed with any directives this encodable requires.
 */
void ProtocolField::getMapIncludeDirectives(std::vector<std::string>& list) const
{
    if(inMemoryType.isStruct)
    {
        const ProtocolStructureModule* struc = parser->lookUpStructure(typeName);

        if(struc != NULL)
            struc->getMapIncludeDirectives(list);

    }// if struct

}// ProtocolField::getMapIncludeDirectives


/*!
 * Return the include directives needed for this encodable's compare functions
 * \param list is push_backed with any directives this encodable requires.
 */
void ProtocolField::getCompareIncludeDirectives(std::vector<std::string>& list) const
{
    if(inMemoryType.isStruct)
    {
        const ProtocolStructureModule* struc = parser->lookUpStructure(typeName);

        if(struc != NULL)
            struc->getCompareIncludeDirectives(list);

    }// if struct

}// ProtocolField::getCompareIncludeDirectives


/*!
 * Return the include directives needed for this encodable's print functions
 * \param list is push_backed with any directives this encodable requires.
 */
void ProtocolField::getPrintIncludeDirectives(std::vector<std::string>& list) const
{
    if(inMemoryType.isStruct)
    {
        const ProtocolStructureModule* struc = parser->lookUpStructure(typeName);

        if(struc != NULL)
            struc->getPrintIncludeDirectives(list);

    }// if struct

}// ProtocolField::getPrintIncludeDirectives


/*!
 * Return the signature of this field in an encode function signature
 * \return The encode signature of this field
 */
std::string ProtocolField::getEncodeSignature(void) const
{
    if(isNotEncoded() || isNotInMemory() || isConstant())
        return "";
    else if(is2dArray())
        return ", const " + typeName + " " + name + "[" + array + "][" + array2d + "]";
    else if(isArray())
        return ", const " + typeName + " " + name + "[" + array + "]";
    else if(!inMemoryType.isStruct)
        return ", " + typeName + " " + name;
    else
        return ", const " + typeName + "* " + name;
}


/*!
 * Get details needed to produce documentation for this encodable.
 * \param outline gives the paragraph numbers for all the fields up to this one.
 * \param startByte is the starting byte location of this encodable, which will be updated for the following encodable.
 * \param bytes is push_backed for the byte range of this encodable.
 * \param names is push_backed for the name of this encodable.
 * \param encodings is push_backed for the encoding of this encodable.
 * \param repeats is push_backed for the array information of this encodable.
 * \param comments is push_backed for the description of this encodable.
 */
void ProtocolField::getDocumentationDetails(std::vector<int>& outline, std::string& startByte, std::vector<std::string>& bytes, std::vector<std::string>& names, std::vector<std::string>& encodings, std::vector<std::string>& repeats, std::vector<std::string>& comments) const
{
    std::string description;
    std::string maxEncodedLength = encodedLength.maxEncodedLength;

    if(encodedType.isNull || hidden)
        return;

    // See if we can replace any enumeration names with values
    maxEncodedLength = parser->replaceEnumerationNameWithValue(maxEncodedLength);

    // The byte after this one
    std::string nextStartByte = EncodedLength::collapseLengthString(startByte + "+" + maxEncodedLength);

    // The length data
    if(encodedType.isBitfield)
    {
        std::string range;

        // the starting bit count is the full count, not the count in the byte
        int startCount = bitfieldData.startingBitCount % 8;

        if(startByte.empty())
            range = "0:" + std::to_string(7 - startCount);
        else
            range = startByte + ":" + std::to_string(7 - startCount);

        if(encodedType.bits > 1)
        {
            int endCount = (startCount + encodedType.bits - 1);

            int byteCount = endCount / 8;

            std::string endByte = EncodedLength::collapseLengthString(startByte + "+" + std::to_string(byteCount), true);

            range += "..." + endByte + ":" + std::to_string(7 - (endCount%8));
        }

        bytes.push_back(range);
    }
    else
    {
        if(maxEncodedLength.empty() || (maxEncodedLength.compare("1") == 0))
            bytes.push_back(startByte);
        else
        {
            std::string endByte = EncodedLength::subtractOneFromLengthString(nextStartByte);

            // The range of the data
            bytes.push_back(startByte + "..." + endByte);
        }
    }

    // The name information
    outline.back() += 1;
    std::string outlineString = std::to_string(outline.at(0));
    for(std::size_t i = 1; i < outline.size(); i++)
        outlineString += "." + std::to_string(outline.at(i));

    if(inMemoryType.isEnum)
    {
        // Link to the enumeration
        outlineString += ")[" + title +"](#" + inMemoryType.enumName + ")";
    }
    else
        outlineString += ")" + title;

    names.push_back(outlineString);

    if(inMemoryType.isStruct)
    {
        // Encoding is blank for structures
        encodings.push_back(std::string());

        // Repeats
        repeats.push_back(getRepeatsDocumentationDetails());

        // Fourth column is the commenting
        description += comment;

        if(!dependsOn.empty())
        {
            if(!endsWith(description, "."))
                description += ".";

            if(dependsOnValue.empty())
                description += " Only included if " + dependsOn + " is non-zero.";
            else
            {
                if(dependsOnCompare.empty())
                    description += " Only included if " + dependsOn + " equal to " + dependsOnValue + ".";
                else
                    description += " Only included if " + dependsOn + " " + dependsOnCompare + " " + dependsOnValue + ".";
            }
        }

        if(description.empty())
            comments.push_back(std::string());
        else
            comments.push_back(description);

        std::string subStartByte = startByte;

        // Now go get the substructure stuff
        parser->getStructureSubDocumentationDetails(typeName, outline, subStartByte, bytes, names, encodings, repeats, comments);

    }// if structure
    else
    {
        // The encoding
        if(encodedType.isFixedString)
        {
            encodings.push_back("Zero terminated string of exactly " + array + " bytes");
            repeats.push_back(std::string());
        }
        else if( encodedType.isString)
        {
            encodings.push_back("Zero-terminated string up to " + array + " bytes");
            repeats.push_back(std::string());
        }
        else
        {
            if(encodedType.isBitfield)
                encodings.push_back("B" + std::to_string(encodedType.bits));
            else if(encodedType.isFloat)
            {
                if(encodedType.bits < 32)
                    encodings.push_back("F" + std::to_string(encodedType.bits) + ":" + std::to_string(encodedType.sigbits));
                else
                    encodings.push_back("F" + std::to_string(encodedType.bits));
            }
            else if(encodedType.isSigned)
                encodings.push_back("I" + std::to_string(encodedType.bits));
            else
                encodings.push_back("U" + std::to_string(encodedType.bits));

            // Third column is the repeat/array column
            repeats.push_back(getRepeatsDocumentationDetails());
        }

        // Fourth column is the commenting
        if(inMemoryType.isNull)
        {
            if(!comment.empty())
                description += comment;
            else
            {
                if(encodedType.isBitfield)
                    description += "Reserved bits in the packet.";
                else
                    description += "Reserved bytes in the packet.";
            }
        }
        else
            description += comment;

        if(!description.empty() && !endsWith(description, "."))
            description += ".";

        if(support.limitonencode && (!verifyMinStringForDisplay.empty() || !verifyMaxStringForDisplay.empty()))
        {
            if(!verifyMinStringForDisplay.empty() && !verifyMaxStringForDisplay.empty())
                description += "<br>Value is limited on encode from " + verifyMinStringForDisplay + " to " + verifyMaxStringForDisplay + ".";
            else if(verifyMinStringForDisplay.empty())
                description += "<br>Value is limited on encode to be less than or equal to " + verifyMaxStringForDisplay + ".";
            else
                description += "<br>Value is limited on encode to be greater than or equal to " + verifyMinStringForDisplay + ".";
        }

        if(isFloatScaling() || isIntegerScaling())
            description += "<br>Scaled by " + scalerString + " from " + getDisplayNumberString(encodedMin) + " to " + getDisplayNumberString(encodedMax) + ".";
        else if(!scalerString.empty() && (encodedType.isFloat))
            description += "<br>Scaled by " + scalerString + ".";

        if(!constantString.empty())
            description += "<br>Data are given constant value on encode " + constantStringForDisplay + ".";

        if(!dependsOn.empty())
        {
            if(dependsOnValue.empty())
                description += "<br>Only included if " + dependsOn + " is non-zero.";
            else
            {
                if(dependsOnCompare.empty())
                    description += "<br>Only included if " + dependsOn + " equal to " + dependsOnValue + ".";
                else
                    description += "<br>Only included if " + dependsOn + " " + dependsOnCompare + " " + dependsOnValue + ".";
            }
        }

        if(!defaultString.empty())
            description += "<br>This field is optional. If it is not included then the value is assumed to be " + defaultStringForDisplay + ".";

        if(overridesPrevious)
            description += "<br>This field overrides the previous field of the same name, if the packet is long enough.";

        for (std::size_t i = 0; i < extraInfoNames.size(); i++)
        {
            if ((extraInfoValues.size() > i) && (!extraInfoValues.at(i).empty()))
                description += "<br>" + extraInfoNames.at(i) + ": " + extraInfoValues.at(i) + ".";
        }

        // StringList cannot be empty
        if(description.empty())
            comments.push_back(std::string());
        else
            comments.push_back(description);

    }// else if not structure

    // Update startByte for following encodables
    startByte = nextStartByte;

}// ProtocolField::getDocumentationDetails


/*!
 * Get the next lines(s) of source coded needed to encode this field
 * \param isBigEndian should be true for big endian encoding.
 * \param bitcount points to the running count of bits in a bitfields and should persist between calls
 * \param isStructureMember should be true if the left hand side is a
 *        member of a user structure, else the left hand side is a pointer
 *        to the inMemoryType
 * \param enforcelimit should be true to perform limiting (if verify limits are provided) as part of the encode.
 * \return The string to add to the source file that encodes this field.
 */
std::string ProtocolField::getEncodeString(bool isBigEndian, int* bitcount, bool isStructureMember) const
{
    std::string output;

    if(encodedType.isBitfield)
    {
        output = getEncodeStringForBitfield(bitcount, isStructureMember);
    }
    else
    {
        if(inMemoryType.isString)
            output += getEncodeStringForString(isStructureMember);
        else if(inMemoryType.isStruct)
            output += getEncodeStringForStructure(isStructureMember);
        else
            output += getEncodeStringForField(isBigEndian, isStructureMember);
    }

    return output;

}


/*!
 * Get the next lines(s) of source coded needed to decode this field
 * \param isBigEndian should be true for big endian encoding.
 * \param bitcount points to the running count of bits in a bitfields and should persist between calls
 * \param isStructureMember should be true if the left hand side is a
 *        member of a user structure, else the left hand side is a pointer
 *        to the inMemoryType
 * \param defaultEnabled should be true to handle defaults
 * \return The string to add to the source file that encodes this field.
 */
std::string ProtocolField::getDecodeString(bool isBigEndian, int* bitcount, bool isStructureMember, bool defaultEnabled) const
{
    std::string output;

    if(encodedType.isBitfield)
        output += getDecodeStringForBitfield(bitcount, isStructureMember, defaultEnabled);
    else if(inMemoryType.isString)
        output += getDecodeStringForString(isStructureMember, defaultEnabled);
    else if(inMemoryType.isStruct)
        output += getDecodeStringForStructure(isStructureMember);
    else
        output += getDecodeStringForField(isBigEndian, isStructureMember, defaultEnabled);

    return output;
}


//! True if this encodable has verification data
bool ProtocolField::hasVerify(void) const
{
    if(inMemoryType.isStruct)
    {
        const ProtocolStructure* struc = parser->lookUpStructure(typeName);
        if(struc)
            return struc->hasVerify();
        else
            return false;
    }
    else
    {
        if(verifyMaxString.empty() && verifyMinString.empty())
            return false;
        else
            return true;
    }

}// ProtocolField::hasVerify


//! True if this encodable has initialization data
bool ProtocolField::hasInit(void) const
{
    if(inMemoryType.isStruct)
    {
        const ProtocolStructure* struc = parser->lookUpStructure(typeName);
        if(struc)
            return struc->hasInit();
        else
            return false;
    }
    else
    {
        if(initialValueString.empty())
            return false;
        else
            return true;
    }

}// ProtocolField::hasInit


/*!
 * Get the string used for verifying this field. If there is no verification data the string will be empty.
 * \return the string used for verifying this field, which may be blank
 */
std::string ProtocolField::getVerifyString(void) const
{
    if(!hasVerify())
        return std::string();

    // No verify for null or string
    if(inMemoryType.isNull || inMemoryType.isString)
        return std::string();

    std::string access;
    std::string spacing = TAB_IN;
    std::string arrayspacing;
    std::string output;

    if(!comment.empty())
        output += spacing + "// " + comment + "\n";

    output += getEncodeArrayIterationCode(spacing, true);

    if(inMemoryType.isStruct)
    {
        if(support.language == ProtocolSupport::c_language)
            access = "&_pg_user->" + name;
        else
            access = name;  // in this case, name is already pointer, so we don't need "&"

        if(isArray())
        {
            arrayspacing += TAB_IN;

            if(support.language == ProtocolSupport::c_language)
                access = "&_pg_user->" + name + "[_pg_i]";
            else
                access = name + "[_pg_i]";

            // Handle 2D array
            if(is2dArray())
            {
                access += "[_pg_j]";
                arrayspacing += TAB_IN;
            }
        }

        // This calls the verify function
        if(support.language == ProtocolSupport::c_language)
        {
            output += spacing + arrayspacing + "if(!verify" + typeName + "(" + access + "))\n";
            output += spacing + arrayspacing + TAB_IN + "_pg_good = 0;\n";
        }
        else
        {
            output += spacing + arrayspacing + "if(!" + access + ".verify())\n";
            output += spacing + arrayspacing + TAB_IN + "_pg_good = false;\n";
        }

    }// if struct being verified
    else
    {
        std::string failedvalue = "false";
        if(support.language == ProtocolSupport::c_language)
            failedvalue = "0";

        if(support.language == ProtocolSupport::c_language)
            access = "_pg_user->" + name;
        else
            access = name;

        if(isArray())
        {
            arrayspacing += TAB_IN;

            if(support.language == ProtocolSupport::c_language)
                access = "_pg_user->" + name + "[_pg_i]";
            else
                access = name + "[_pg_i]";

            // Handle 2D array
            if(is2dArray())
            {
                access += "[_pg_j]";
                arrayspacing += TAB_IN;
            }
        }

        if(!verifyMinString.empty())
        {
            output += spacing + arrayspacing + "if(" + access + " < " + verifyMinString + ")\n";
            output += spacing + arrayspacing + "{\n";
            output += spacing + arrayspacing + TAB_IN + access + " = " + verifyMinString + ";\n";
            output += spacing + arrayspacing + TAB_IN + "_pg_good = " + failedvalue + ";\n";
            output += spacing + arrayspacing + "}\n";
        }

        if(!verifyMaxString.empty())
        {
            std::string choice = "else if(";
            if(verifyMinString.empty())
                choice = "if(";

            output += spacing + arrayspacing + choice + access + " > " + verifyMaxString + ")\n";
            output += spacing + arrayspacing + "{\n";
            output += spacing + arrayspacing + TAB_IN + access + " = " + verifyMaxString + ";\n";
            output += spacing + arrayspacing + TAB_IN + "_pg_good = " + failedvalue + ";\n";
            output += spacing + arrayspacing + "}\n";
        }

    }// else if not struct

    return output;

}// ProtocolField::getVerifyString


/*!
* Get the string used for comparing this field.
* \return the string used to compare this field, which may be empty
*/
std::string ProtocolField::getComparisonString(void) const
{
    std::string output;

    // No comparison if nothing is in memory or if not encoded
    if(inMemoryType.isNull || encodedType.isNull)
        return output;

    if(inMemoryType.isString)
    {
        if(support.language == ProtocolSupport::c_language)
            output += TAB_IN + "if(std::string(_pg_user1->" + name + ").compare(_pg_user2->" + name + ") != 0)\n";
        else
            output += TAB_IN + "if(std::string(" + name + ").compare(_pg_user->" + name + ") != 0)\n";

        output += TAB_IN + TAB_IN + "_pg_report += _pg_prename + \":" + name + " strings differ\\n\";\n";
    }
    else
    {
        std::string spacing = TAB_IN;
        bool closearraytest = false;
        bool closearraytest2 = false;
        bool closeforloop = false;
        bool closeforloop2 = false;

        std::string access1, access2;
        if(support.language == ProtocolSupport::c_language)
        {
            access1 = "_pg_user1->" + name;
            access2 = "_pg_user2->" + name;
        }
        else
        {
            access1 = name;
            access2 = "_pg_user->" + name;
        }


        if(isArray())
        {
            if(!variableArray.empty())
            {
                if(support.language == ProtocolSupport::c_language)
                    output += spacing + "if(_pg_user1->" + variableArray + " == _pg_user2->" + variableArray + ")\n";
                else
                    output += spacing + "if(" + variableArray + " == _pg_user->" + variableArray + ")\n";

                output += spacing + "{\n";
                spacing += TAB_IN;
                closearraytest = true;

                if(support.language == ProtocolSupport::c_language)
                    output += spacing + "for(_pg_i = 0; (_pg_i < " + array + ") && (_pg_i < (unsigned)_pg_user1->" + variableArray + "); _pg_i++)\n";
                else
                    output += spacing + "for(_pg_i = 0; (_pg_i < " + array + ") && (_pg_i < (unsigned)" + variableArray + "); _pg_i++)\n";

                output += spacing + "{\n";
                spacing += TAB_IN;
                closeforloop = true;
            }
            else
            {
                output += spacing + "for(_pg_i = 0; _pg_i < " + array + "; _pg_i++)\n";
                spacing += TAB_IN;
            }

            access1 += "[_pg_i]";
            access2 += "[_pg_i]";
        }

        if(is2dArray())
        {
            if(!variable2dArray.empty())
            {
                if(support.language == ProtocolSupport::c_language)
                    output += spacing + "if(_pg_user1->" + variable2dArray + " == _pg_user2->" + variable2dArray + ")\n";
                else
                    output += spacing + "if(" + variable2dArray + " == _pg_user->" + variable2dArray + ")\n";

                output += spacing + "{\n";
                spacing += TAB_IN;
                closearraytest2 = true;

                if(support.language == ProtocolSupport::c_language)
                    output += spacing + "for(_pg_j = 0; (_pg_j < " + array2d + ") && (_pg_j < (unsigned)_pg_user1->" + variable2dArray + "); _pg_j++)\n";
                else
                    output += spacing + "for(_pg_j = 0; (_pg_j < " + array2d + ") && (_pg_j < (unsigned)" + variable2dArray + "); _pg_j++)\n";

                output += spacing + "{\n";
                spacing += TAB_IN;
                closeforloop2 = true;
            }
            else
            {
                output += spacing + "for(_pg_j = 0; _pg_j < " + array2d + "; _pg_j++)\n";
                spacing += TAB_IN;
            }

            access1 += "[_pg_j]";
            access2 += "[_pg_j]";
        }

        if(inMemoryType.isStruct)
        {
            if(support.language == ProtocolSupport::c_language)
                output += spacing + "_pg_report += compare" + typeName + "(_pg_prename + \":" + name + "\"";
            else
                output += spacing + "_pg_report += " + access1 + ".compare(_pg_prename + \":" + name + "\"";

            if(isArray())
                output += " + \"[\" + std::to_string(_pg_i) + \"]\"";

            if(is2dArray())
                output += " + \"[\" + std::to_string(_pg_j) + \"]\"";

            // Structure compare we need to pass the address of the structure, not the object
            if(support.language == ProtocolSupport::c_language)
                output += ", &" + access1 + ", &" + access2 + ");\n";
            else
                output += ", &" + access2 + ");\n";
        }
        else
        {
            // Simple value comparison to generate the _pg_report
            output += spacing + "if(" + access1 + " != " + access2 + ")\n";

            // The _pg_report includes the _pg_prename and the name
            output += spacing + TAB_IN + "_pg_report += _pg_prename + \":" + name + "\" ";

            // The _pg_report needs to include the array indices
            if(isArray())
                output += " + \"[\" + std::to_string(_pg_i) + \"]\"";

            if(is2dArray())
                output += " + \"[\" + std::to_string(_pg_j) + \"]\"";

            // And finally the values go into the _pg_report
            if(!printScalerString.empty())
                output += " + \" '\" + to_formatted_string(" + access1 + printScalerString + ", 16) + \"' '\" + to_formatted_string(" + access2 + printScalerString + ", 16) + \"'\\n\";\n";
            else if(inMemoryType.isFloat && (inMemoryType.bits > 32))
                output += " + \" '\" + to_formatted_string(" + access1 + ", 16) + \"' '\" + to_formatted_string(" + access2 + ", 16) + \"'\\n\";\n";
            else if(inMemoryType.isFloat)
                output += " + \" '\" + to_formatted_string(" + access1 + ", 7) + \"' '\" + to_formatted_string(" + access2 + ", 7) + \"'\\n\";\n";
            else
                output += " + \" '\" + std::to_string(" + access1 + ") + \"' '\" + std::to_string(" + access2 + ") + \"'\\n\";\n";

        }// else not a struct

        if(closeforloop2)
        {
            spacing.erase(spacing.size() - TAB_IN.size(), TAB_IN.size());
            output += spacing + "}\n";
        }

        if(closearraytest2)
        {
            spacing.erase(spacing.size() - TAB_IN.size(), TAB_IN.size());
            output += spacing + "}\n";
            output += spacing + "else\n";
            output += spacing + TAB_IN + "_pg_report += _pg_prename + \":" + name + " 2nd array dimension differs, array not compared\\n\";\n";
        }

        if(closeforloop)
        {
            spacing.erase(spacing.size() - TAB_IN.size(), TAB_IN.size());
            output += spacing + "}\n";
        }

        if(closearraytest)
        {
            spacing.erase(spacing.size() - TAB_IN.size(), TAB_IN.size());
            output += spacing + "}\n";
            output += spacing + "else\n";
            output += spacing + TAB_IN + "_pg_report += _pg_prename + \":" + name + " array dimension differs, array not compared\\n\";\n";
        }

    }// else numeric output

    return output;

}// ProtocolField::getComparisonString


/*!
 * Get the string used for text printing this field.
 * \return the string used to print this field as text, which may be empty
 */
std::string ProtocolField::getTextPrintString(void) const
{
    std::string output;

    // No print if nothing is in memory or if not encoded
    if(inMemoryType.isNull || encodedType.isNull)
        return output;

    if(inMemoryType.isString)
    {
        output += TAB_IN + "_pg_report += _pg_prename + \":" + name + " '\" + std::string(" + getEncodeFieldAccess(true) + ") + \"'\\n\";\n";
    }
    else
    {
        std::string spacing = TAB_IN;

        output += getEncodeArrayIterationCode(spacing, true);
        if(isArray())
        {
            spacing += TAB_IN;
            if(is2dArray())
                spacing += TAB_IN;
        }

        if(inMemoryType.isStruct)
        {
            if(support.language == ProtocolSupport::c_language)
                output += spacing + "_pg_report += textPrint" + typeName + "(_pg_prename + \":" + name + "\"";
            else
                output += spacing + "_pg_report += " + getEncodeFieldAccess(true) + ".textPrint(_pg_prename + \":" + name + "\"";

            if(isArray())
                output += " + \"[\" + std::to_string(_pg_i) + \"]\"";

            if(is2dArray())
                output += " + \"[\" + std::to_string(_pg_j) + \"]\"";

            if(support.language == ProtocolSupport::c_language)
                output += ", " + getEncodeFieldAccess(true);

            output += ");\n";
        }
        else
        {
            // The _pg_report includes the _pg_prename and the name
            output += spacing + "_pg_report += _pg_prename + \":" + name + "\"";

            // The _pg_report needs to include the array indices
            if(isArray())
                output += " + \"[\" + std::to_string(_pg_i) + \"]\"";

            if(is2dArray())
                output += " + \"[\" + std::to_string(_pg_j) + \"]\"";

            // And finally the values go into the _pg_report
            if(!printScalerString.empty())
                output += " + \" '\" +  to_formatted_string(" + getEncodeFieldAccess(true) + printScalerString + ", 16) + \"'\\n\";\n";
            else if(inMemoryType.isFloat && (inMemoryType.bits > 32))
                output += " + \" '\" +  to_formatted_string(" + getEncodeFieldAccess(true) + ", 16) + \"'\\n\";\n";
            else if(inMemoryType.isFloat)
                output += " + \" '\" +  to_formatted_string(" + getEncodeFieldAccess(true) + ", 7) + \"'\\n\";\n";
            else
                output += " + \" '\" + std::to_string(" + getEncodeFieldAccess(true) + ") + \"'\\n\";\n";

        }// else not a struct

    }// else numeric output

    return output;

}// ProtocolField::getTextPrintString


/*!
 * Get the string used for text reading this field.
 * \return the string used to read this field as text, which may be empty
 */
std::string ProtocolField::getTextReadString(void) const
{
    std::string output;

    // No print if nothing is in memory or if not encoded
    if(inMemoryType.isNull || encodedType.isNull)
        return output;

    if(inMemoryType.isString)
    {
        // Notice the use of the "pg" copy function, which is just like strncpy but without the security vulnerabilities
        output += TAB_IN + "pgstrncpy(" + getDecodeFieldAccess(true) + ", extractText(_pg_prename + \":" + name + "\", _pg_source, &_pg_fieldcount).c_str(), " + array + ");\n";
    }
    else
    {
        std::string spacing = TAB_IN;

        output += getDecodeArrayIterationCode(spacing, true);

        if(is2dArray())
        {
            spacing += TAB_IN;
            output += spacing + "{\n";
            spacing += TAB_IN;
        }
        else if(isArray())
        {
            output += spacing + "{\n";
            spacing += TAB_IN;
        }

        if(inMemoryType.isStruct)
        {
            if(support.language == ProtocolSupport::c_language)
                output += spacing + "_pg_fieldcount += textRead" + typeName + "(_pg_prename + \":" + name + "\"";
            else
                output += spacing + "_pg_fieldcount += " + getDecodeFieldAccess(true) + ".textRead(_pg_prename + \":" + name + "\"";

            if(isArray())
                output += " + \"[\" + std::to_string(_pg_i) + \"]\"";

            if(is2dArray())
                output += " + \"[\" + std::to_string(_pg_j) + \"]\"";

            // Structure read, we need to pass the address of the structure, not the object        
            if(support.language == ProtocolSupport::c_language)
                output += ", _pg_source, " + getDecodeFieldAccess(true) + ");\n";
            else
                output += ", _pg_source);\n";
        }
        else
        {
            // First get the text
            output += spacing + "_pg_text = extractText(_pg_prename + \":" + name + "\"";

            // The array indices are part of the key text
            if(isArray())
            {
                output += " + \"[\" + std::to_string(_pg_i) + \"]\"";

                if(is2dArray())
                    output += " + \"[\" + std::to_string(_pg_j) + \"]\"";
            }

            output += ", _pg_source, &_pg_fieldcount);\n";

            // Check the text and get a result if it is not empty
            output += spacing + "if(!_pg_text.empty())\n";

            if(!readScalerString.empty())
                output += spacing + TAB_IN + getDecodeFieldAccess(true) + " = (" + typeName + ")(std::stod(_pg_text)" + readScalerString + ");\n";
            else if(inMemoryType.isFloat && (inMemoryType.bits > 32))
                output += spacing + TAB_IN + getDecodeFieldAccess(true) + " = std::stod(_pg_text);\n";
            else if(inMemoryType.isFloat)
                output += spacing + TAB_IN + getDecodeFieldAccess(true) + " = std::stof(_pg_text);\n";
            else if(inMemoryType.isSigned)
            {
                if(inMemoryType.bits > 32)
                    output += spacing + TAB_IN + getDecodeFieldAccess(true) + " = (" + typeName + ")(std::stoll(_pg_text));\n";
                if(inMemoryType.bits > 16)
                    output += spacing + TAB_IN + getDecodeFieldAccess(true) + " = (" + typeName + ")(std::stol(_pg_text));\n";
                else
                    output += spacing + TAB_IN + getDecodeFieldAccess(true) + " = (" + typeName + ")(std::stoi(_pg_text));\n";
            }
            else
            {
                if(inMemoryType.bits > 32)
                    output += spacing + TAB_IN + getDecodeFieldAccess(true) + " = (" + typeName + ")(std::stoull(_pg_text));\n";
                else
                    output += spacing + TAB_IN + getDecodeFieldAccess(true) + " = (" + typeName + ")(std::stoul(_pg_text));\n";
            }

        }// else not a struct

        // Close the block under the for loop(s)
        if(isArray())
        {
            spacing = spacing.substr(0, spacing.length() - TAB_IN.length());
            output += spacing + "}\n";
        }

    }// else numeric output

    return output;

}// ProtocolField::getTextReadString


/*!
 * Get the string used for storing this field in a map.
 * \return the string used to read this field as text, which may be empty
 */
std::string ProtocolField::getMapEncodeString(void) const
{  
    std::string output;

    // Cannot encode to a null memory type
    if(inMemoryType.isNull)
        return output;

    // Return empty string if this field should not be encoded to map
    if((mapOptions & MAP_ENCODE) == 0)
        return output;

    if(inMemoryType.isNull || encodedType.isNull)
        return output;

    if(!comment.empty())
        output += TAB_IN + "// " + comment + "\n";

    if(inMemoryType.isString)
    {
        output += TAB_IN + "_pg_map[_pg_prename + \":" + name + "\"] = QString(" + getEncodeFieldAccess(true) + ");\n";
    }
    else
    {
        std::string spacing = TAB_IN;

        output += getEncodeArrayIterationCode(spacing, true);
        if(isArray())
        {
            spacing += TAB_IN;
            if(is2dArray())
                spacing += TAB_IN;
        }

        if(inMemoryType.isStruct)
        {
            if(support.language == ProtocolSupport::c_language)
                output += spacing + "mapEncode" + typeName + "(_pg_prename + \":" + name + "\"";
            else
                output += spacing + getEncodeFieldAccess(true) + ".mapEncode(_pg_prename + \":" + name + "\"";

            if(isArray())
                output += " + \"[\" + QString::number(_pg_i) + \"]\"";
            if(is2dArray())
                output += " + \"[\" + QString::number(_pg_j) + \"]\"";

            if(support.language == ProtocolSupport::c_language)
                output += ", _pg_map, " + getEncodeFieldAccess(true) + ");\n";
            else
                output += ", _pg_map);\n";

        }// data type is a struct
        else
        {
            output += spacing + "_pg_map[_pg_prename + \":" + name + "\"";

            if(isArray())
                output += " + \"[\" + QString::number(_pg_i) + \"]\"";
            if(is2dArray())
                output += " + \"[\" + QString::number(_pg_j) + \"]\"";

            output += "] = ";

            // Numeric values are automatically converted to correct QVariant types
            if(inMemoryType.isFloat || !printScalerString.empty())
                output += getEncodeFieldAccess(true) + printScalerString;
            else if (inMemoryType.isBool && (support.language == ProtocolSupport::c_language || support.language == ProtocolSupport::cpp_language))
            {
                // Ensure that boolean types are encoded as unsigned chars
                output += "(unsigned char) " + getEncodeFieldAccess(true);
            }
            else
                output += getEncodeFieldAccess(true);

            output += ";\n";

        }// else not a struct

    }// else numeric output

    return output;

}// ProtocolField::getMapEncodeString


/*!
 * Get the string used for extracting this field from a map.
 * \return the string used to read this field as text, which may be empty
 */
std::string ProtocolField::getMapDecodeString(void) const
{
    std::string key;
    std::string output;
    std::string decode;

    // Cannot decode to a null memory type
    if(inMemoryType.isNull)
        return output;

    // Return empty string if this field is not to be decoded
    if((mapOptions & MAP_DECODE) == 0)
        return output;

    if(!comment.empty())
        output += TAB_IN + "// " + comment + "\n";

    if(inMemoryType.isString)
    {
        key = "_pg_prename + \":" + name + "\"";

        output += TAB_IN + "key = " + key + ";\n";
        output += "\n";
        output += TAB_IN + "if (_pg_map.contains(key))\n";
        output += TAB_IN + TAB_IN + "qstrncpy(" + getDecodeFieldAccess(true) + ", _pg_map[key].toString().toLatin1().constData(), " + array + ");\n";
    }
    else
    {
        std::string spacing = TAB_IN;

        output += getEncodeArrayIterationCode(spacing, true);

        if(is2dArray())
        {
            spacing += TAB_IN;
            output += spacing + "{\n";
            spacing += TAB_IN;
        }
        else if(isArray())
        {
            output += spacing + "{\n";
            spacing += TAB_IN;
        }

        if(inMemoryType.isStruct)
        {
            if(support.language == ProtocolSupport::c_language)
                output += spacing + "mapDecode" + typeName + "(_pg_prename + \":" + name + "\"";
            else
                output += spacing + getDecodeFieldAccess(true) + ".mapDecode(_pg_prename + \":" + name + "\"";

            if(isArray())
                output += " + \"[\" + QString::number(_pg_i) + \"]\"";
            if(is2dArray())
                output += " + \"[\" + QString::number(_pg_j) + \"]\"";

            if(support.language == ProtocolSupport::c_language)
                output += ", _pg_map, " + getDecodeFieldAccess(true) + ");\n";
            else
                output += ", _pg_map);\n";

        }// data type is a struct
        else
        {
            key = "_pg_prename + \":" + name + "\"";

            if(isArray())
                key += " + \"[\" + QString::number(_pg_i) + \"]\"";
            if(is2dArray())
                key += " + \"[\" + QString::number(_pg_j) + \"]\"";

            output += spacing + "key = " + key + ";\n";

            decode = "_pg_map[key]";

            if(inMemoryType.isFloat || !printScalerString.empty())
                decode += ".toDouble";
            else if (inMemoryType.isSigned)
            {
                if(inMemoryType.bits > 32)
                    decode += ".toLongLong";
                else
                    decode += ".toInt";
            }
            else
            {
                if(inMemoryType.bits > 32)
                    decode += ".toULongLong";
                else
                    decode += ".toUInt";
            }

            /* Generate code to load the data from the map:
             * 1. Check that the given key exists
             * 2. Check that the provided value decodes properly (for the given data type)
             * 3. Load out the value from the map if 1. and 2. pass
             */

            output += spacing + "if(_pg_map.contains(key))\n";
            output += spacing + "{\n";
            output += spacing + TAB_IN + decode + "(&ok);\n";
            output += spacing + TAB_IN + "if(ok)\n";
            output += spacing + TAB_IN + TAB_IN + getDecodeFieldAccess(true) + " = (" + typeName + ")" + decode + "()";
            if(inMemoryType.isFloat && !printScalerString.empty())
                output += readScalerString;
            output += ";\n";
            output += spacing + "}\n";

        }// else not a struct

        // Close the block under the for loop(s)
        if(isArray())
        {
            spacing = spacing.substr(0, spacing.length() - TAB_IN.length());
            output += spacing + "}\n";
        }

    }// else numeric output

    return output;

}// ProtocolField::getMapDecodeString


/*!
 * Return the string that sets this encodable to its default value in code
 * \param isStructureMember should be true if this field is accessed through a "user" structure pointer
 * \return the string to add to the source file, including line feed
 */
std::string ProtocolField::getSetToDefaultsString(bool isStructureMember) const
{
    return getSetToValueString(isStructureMember, defaultString);
}


/*!
 * Return the string that sets this encodable to its initial value in code
 * \param isStructureMember should be true if this field is accessed through a "user" structure pointer
 * \return the string to add to the source file, including line feed
 */
std::string ProtocolField::getSetInitialValueString(bool isStructureMember) const
{
    std::string output;

    if(inMemoryType.isNull)
        return output;

    if(inMemoryType.isStruct)
    {
        // In C++ the constructor is the initializer
        if(support.language == ProtocolSupport::c_language)
        {
            if(!hasInit())
                return output;

            std::string spacing = TAB_IN;

            if(!comment.empty())
                output += spacing + "// " + comment + "\n";

            getEncodeArrayIterationCode(spacing, isStructureMember);

            if(isArray())
            {
                spacing += TAB_IN;
                if(is2dArray())
                    spacing += TAB_IN;
            }

            output += spacing + "init" + typeName + "(" + getDecodeFieldAccess(isStructureMember) + ");\n";

        }// If C language

    }// if struct
    else if(!isNotInMemory())
    {
        if(support.language == ProtocolSupport::c_language)
        {            
            if(!initialValueString.empty())
            {
                if(!comment.empty())
                    output += TAB_IN + "// " + comment + "\n";

                // This has the array handling built in
                output += getSetToValueString(isStructureMember, initialValueString);
            }
        }
        else
        {
            // Try the user's value first
            std::string initial = initialValueString;

            // If there isn't one, use the default value
            if(initial.empty())
                initial = defaultString;

            // If there isn't one, use the constant value
            if(initial.empty())
                initial = constantString;

            // If there isn't one, use the verify min value
            if(initial.empty())
                initial = verifyMinString;

            // In C++ we explicitly initialize all members.
            if(initial.empty())
            {
                if(!inMemoryType.isString)
                {
                    if(inMemoryType.isEnum)
                    {
                        const EnumCreator* creator = parser->lookUpEnumeration(inMemoryType.enumName);
                        if(creator == nullptr)
                            initial = "(" + inMemoryType.enumName + ")0";
                        else
                            initial = creator->getFirstEnumerationName();
                    }
                    else
                    {
                        // Zero seems like the best choice, but we can do a
                        // little better, if for example we have a minimum
                        // encoded value we should initialize to
                        // respect those values
                        if(limitMaxValue < 0)
                            initial = limitMaxString;
                        else if(limitMinValue > 0)
                            initial = limitMinString;
                        else
                            initial = "0";

                    }// else if we are not an enumeration

                }// If we are not a string

            }// If we have no user provided initial value

            initial = inMemoryType.applyTypeToConstant(initial);

            // C++ initializer list
            if(inMemoryType.isString)
            {
                // initial is a string literal, so include the quotes. Except for
                // a special case. If initial ends in "()" then we assume its a
                // function or macro call
                if(!endsWith(trimm(initial), "()"))
                    initial = "\"" + initial + "\"";

                output += TAB_IN + name + "(" + initial + "),\n";
            }
            else if(is2dArray())
                output += TAB_IN + name + "{{" + initial + "}},\n";
            else if(isArray())
                output += TAB_IN + name + "{" + initial + "},\n";
            else
                output += TAB_IN + name + "(" + initial + "),\n";
        }
    }

    return output;

}// ProtocolField::getSetInitialValueString


/*!
 * Return the string that sets this encodable to specific value in code. Note
 * that for arrays this sets all members, even if the array is variable length.
 * \param isStructureMember should be true if this field is accessed through a
 *        "user" structure pointer.
 * \param value is the string of the specific value.
 * \return the string to add to the source file, including line feed.
 */
std::string ProtocolField::getSetToValueString(bool isStructureMember, std::string value) const
{
    std::string output;

    if(inMemoryType.isStruct)
        return output;

    if(value.empty())
        return output;

    std::string access = getDecodeFieldAccess(isStructureMember);

    // Write out the defaults code
    if(inMemoryType.isString)
    {
        if(value.empty() || (toLower(value) == "null"))
            output += TAB_IN + access + "[0] = 0;\n";
        else
            output += TAB_IN + "pgstrncpy((char*)" + access + ", \"" + value + "\", " + array + ");\n";
    }
    else
    {
        // Deal with casting doubles to floats
        if(inMemoryType.isFloat && (inMemoryType.bits < 64))
        {
            // If the value contains a decimal then we assume it is a floating
            // point constant, if it does not contain ".0f" then we assume it
            // is a double precision constant
            if((value.find('.') != std::string::npos) && (value.back() != 'f'))
            {
                // Rather than presume to change the default constant we simply put a cast in front of it
                value = "(float)" + value;
            }
        }

        if(isArray())
        {
            if(is2dArray())
            {
                output += TAB_IN + "for(_pg_i = 0; _pg_i < " + array + "; _pg_i++)\n";
                output += TAB_IN + TAB_IN + "for(_pg_j = 0; _pg_j < " + array2d + "; _pg_j++)\n";
                output += TAB_IN + TAB_IN + TAB_IN + access + " = " + value + ";\n";
            }
            else
            {
                output += TAB_IN + "for(_pg_i = 0; _pg_i < " + array + "; _pg_i++)\n";
                output += TAB_IN + TAB_IN + access + " = " + value + ";\n";
            }
        }
        else
        {
            output += TAB_IN + access + " = " + value + ";\n";
        }

    }// else if not a struct

    return output;

}// ProtocolField::getSetToValueString


//! Return the strings that #define initial and variable values
std::string ProtocolField::getInitialAndVerifyDefines(bool includeComment) const
{
    (void)includeComment;

    std::string output;

    if(inMemoryType.isNull || inMemoryType.isStruct)
        return output;

    // No colons in preprocessor definitions
    std::string start = getHierarchicalName();
    replaceinplace(start, ":", "_");

    if(!initialValueString.empty())
        output += "#define " + start + "_InitValue " + initialValueString + "\n";

    if(!verifyMinString.empty())
        output += "#define " + start + "_VerifyMin " + verifyMinString + "\n";

    if(!verifyMaxString.empty())
        output += "#define " + start + "_VerifyMax " + verifyMaxString + "\n";

    return output;

}// ProtocolField::getInitialAndVerifyDefines


//! True if this encodable has a direct child that uses bitfields
bool ProtocolField::usesBitfields(void) const
{
    return (encodedType.isBitfield && !isNotEncoded());
}


//! True if this bitfield crosses a byte boundary
bool ProtocolField::bitfieldCrossesByteBoundary(void) const
{
    if(!usesBitfields())
        return false;

    // No byte boundary crossing if only one bit
    if(encodedType.bits <= 1)
        return false;

    // Greater than 8 bits crosses byte boundary for sure
    if(encodedType.bits > 8)
        return true;

    // To check the in-between cases, we have to check the starting bit count
    if(((bitfieldData.startingBitCount % 8) + encodedType.bits) > 8)
        return true;

    return false;
}


//! True if this encodable needs a temporary buffer for its bitfield
bool ProtocolField::usesEncodeTempBitfield(void) const
{
    // We need a temporary bitfield for encode under the following circumstances:
    // 0) Bitfield length is less than 32 bits AND
    // 1) The in-memory type uses more bits than the encoded type, requiring a size check OR
    // 2) The encoded bitfield crosses a byte boundary
    if(usesBitfields() && (encodedType.bits <= 32))
    {
        // If we are encoding a constant zero we don't need temporary variables, no matter how big the field is
        if(getConstantString() == "0")
            return false;

        return bitfieldCrossesByteBoundary();
    }

    // If we get here we do not need a temporary variable for the bitfield
    return false;
}


//! True if this encodable needs a temporary long buffer for its bitfield
bool ProtocolField::usesEncodeTempLongBitfield(void) const
{
    // We need a temporary long bitfield under the following circumstances:
    // 0) Bitfield length is more than 32 bits AND
    // 1) The in-memory type uses more bits than the encoded type, requiring a size check OR
    // 2) The encoded bitfield crosses a byte boundary
    if(usesBitfields() && (encodedType.bits > 32))
    {
        // If we are encoding a constant zero we don't need temporary variables, no matter how big the field is
        if(getConstantString() == "0")
            return false;

        return bitfieldCrossesByteBoundary();
    }

    // If we get here we do not need a temporary variable for the bitfield
    return false;
}


//! True if this encodable needs a temporary buffer for its bitfield
bool ProtocolField::usesDecodeTempBitfield(void) const
{
    // We need a temporary bitfield for decode under the following circumstances:
    // 0) Bitfield length is less than 32 bits AND
    // 1) The in-memory type is null, and we need to do a constant check
    // 2) The encoded bitfield crosses a byte boundary
    if(usesBitfields() && (encodedType.bits <= 32))
    {
        // If we have no in-memory type there is no decoding to do
        if(inMemoryType.isNull)
        {
            // Unless we are checking a constant
            if(checkConstant)
                return true;
            else
                return false;
        }

        // If we are scaling we need a temporary
        if(isFloatScaling() || isIntegerScaling())
            return true;

        return bitfieldCrossesByteBoundary();
    }

    // If we get here we do not need a temporary variable for the bitfield
    return false;
}


//! True if this encodable needs a temporary long buffer for its bitfield
bool ProtocolField::usesDecodeTempLongBitfield(void) const
{
    // We need a temporary long bitfield under the following circumstances:
    // 0) Bitfield length is more than 32 bits AND
    // 1) The in-memory type is null, and we need to do a constant check
    // 2) The encoded bitfield crosses a byte boundary
    if(usesBitfields() && (encodedType.bits > 32))
    {
        // If we have no in-memory type there is no decoding to do
        if(inMemoryType.isNull)
        {
            // Unless we are checking a constant
            if(checkConstant)
                return true;
            else
                return false;
        }

        // If we are scaling we need a temporary
        if(isFloatScaling() || isIntegerScaling())
            return true;

        return bitfieldCrossesByteBoundary();
    }

    // If we get here we do not need a temporary variable for the bitfield
    return false;
}


/*!
 * Create a comment, with no leading white space, but including a line feed,
 * that documents the encoded range of this field. The encoded range will
 * include any modification caused by "limitonencode" rules.
 * \param limitonencode should be true to use the verify min and max strings as part of the range comment
 * \return The range comment
 */
std::string ProtocolField::getRangeComment(bool limitonencode) const
{
    std::string minstring, maxstring;

    if((limitonencode == false) || verifyMaxString.empty() || (hasVerifyMaxValue && (verifyMaxValue >= limitMaxValue)))
        maxstring = limitMaxStringForComment;
    else
        maxstring = verifyMaxString;

    if((limitonencode == false) || verifyMinString.empty() || (hasVerifyMinValue && (verifyMinValue <= limitMinValue)))
        minstring = limitMinStringForComment;
    else
        minstring = verifyMinString;

    return "// Range of " + name + " is " + minstring + " to " + maxstring +  ".\n";

}// ProtocolField::getRangeComment


/*!
 * Given an argument determine if that argument should be passed to the limiting
 * macro and return the argument including the limiting macro as needed.
 * \param argument is the symbol that may need to be limited
 * \return argument embedded within the limiting macro as needed
 */
std::string ProtocolField::getLimitedArgument(std::string argument) const
{
    // Boolean is handled specially, because it is only ever 1 bit (true or false), but the in-memory size is more than 1 bit
    if(inMemoryType.isBool)
        return argument;

    // If scaling is active the scaling functions will apply encoding
    // limits. The same thing will happen with the conversion to smaller
    // floating point types
    if(isFloatScaling() || isIntegerScaling() || (encodedType.isFloat && (encodedType.bits < 32)))
    {
        // However the user may want tighter limits
        if(support.limitonencode && (!verifyMinString.empty() || !verifyMaxString.empty()))
        {
            bool skipmin = verifyMinString.empty();
            bool skipmax = verifyMaxString.empty();

            if(!skipmax && hasVerifyMaxValue && (verifyMaxValue >= limitMaxValue))
                skipmax = true;

            if(!skipmin && hasVerifyMinValue && (verifyMinValue <= limitMinValue))
                skipmin = true;

            if(skipmin && skipmax)
                return argument;

            if(skipmin)
                argument = "limitMax(" + argument + ", " + verifyMaxString + ")";
            else if(skipmax)
                argument = "limitMin(" + argument + ", " + verifyMinString + ")";
            else
                argument = "limitBoth(" + argument + ", " + verifyMinString + ", " + verifyMaxString + ")";

        }// if user asked for specific encode limiting

    }// if scaling is going on
    else
    {
        std::string minstring = limitMinString;
        std::string maxstring = limitMaxString;
        double minvalue = limitMinValue;
        double maxvalue = limitMaxValue;
        bool skipmin = true;
        bool skipmax = true;

        // In this case we don't have the scaling functions, so we may need to apply a limit, even if the user didn't ask for it
        if(!hasVerifyMaxValue && !verifyMaxString.empty() && support.limitonencode)
        {
            // In this case we cannot vet the user's verify string, we just have to use it
            maxstring = verifyMaxString;
            skipmax = false;

        }// if we cannot evaluate the verify value
        else
        {
            if(hasVerifyMaxValue && support.limitonencode && (verifyMaxValue < limitMaxValue))
            {
                maxvalue = verifyMaxValue;
                maxstring = verifyMaxString;
            }

            // Now check if this max value is less than the in-Memory maximum. If it is then we must apply the max limit. We allow one lsb of fiddle.
            if(nextafter(maxvalue, std::numeric_limits<double>::max()) < inMemoryType.getMaximumFloatValue())
                skipmax = false;

        }// else if we can evaluate the verify value

        // In this case we don't have the scaling functions, so we may need to apply a limit, even if the user didn't ask for it
        if(!hasVerifyMinValue && !verifyMinString.empty() && support.limitonencode)
        {
            // In this case we cannot vet the user's verify string, we just have to use it
            minstring = verifyMinString;
            skipmin = false;

        }// if we cannot evaluate the verify value
        else
        {
            if(hasVerifyMinValue && support.limitonencode && (verifyMinValue > limitMinValue))
            {
                minvalue = verifyMinValue;
                minstring = verifyMinString;
            }

            // Now check if this min value is more than the in-Memory minimum. If it is then we must apply the min limit. We allow one lsb of fiddle.
            if(nextafter(minvalue, -std::numeric_limits<double>::max()) > inMemoryType.getMinimumFloatValue())
                skipmin = false;

        }// else if we can evaluate the verify value

        if(skipmin && skipmax)
            return argument;

        if(skipmin)
            argument = "limitMax(" + argument + ", " + maxstring + ")";
        else if(skipmax)
            argument = "limitMin(" + argument + ", " + minstring + ")";
        else
            argument = "limitBoth(" + argument + ", " + minstring + ", " + maxstring + ")";

    }// else if not scaling

    return argument;

}// ProtocolField::getLimitedArgument


/*!
 * Get the next lines(s) of source coded needed to encode this bitfield field
 * \param bitcount points to the running count of bits in this string of
 *        bitfields, and will be updated by this fields encoded bit count.
 * \param isStructureMember should be true if the left hand side is a
 *        member of a user structure, else the left hand side is a pointer
 *        to the inMemoryType
 * \return The string to add to the source file that encodes this field.
 */
std::string ProtocolField::getEncodeStringForBitfield(int* bitcount, bool isStructureMember) const
{
    std::string argument;
    std::string output;
    std::string constantstring = getConstantString();

    if(encodedType.isNull)
        return output;

    if(!comment.empty())
        output += TAB_IN + "// " + comment + "\n";

    // Commenting indicating the range of the field
    if(!inMemoryType.isNull && !inMemoryType.isBool && constantstring.empty() && (encodedType.bits > 1) && !inMemoryType.isEnum)
        output += TAB_IN + getRangeComment(support.limitonencode);

    if(constantstring.empty())
        argument = getLimitedArgument(getEncodeFieldAccess(isStructureMember));
    else
        argument = constantstring;

    if(inMemoryType.isBool)
    {
        argument = "((" + argument + " == true) ? 1 : 0)";
    }
    else if(isFloatScaling())
    {
        if(encodedType.bits > 32)
        {
            if(support.longbitfield)
            {
                if(support.float64)
                    argument = "float64ScaledToLongBitfield((double)" + argument;
                else
                    argument = "float32ScaledToLongBitfield((float)" + argument;
            }
            else
            {
                if(support.float64)
                   argument = "float64ScaledToBitfield((double)" + argument;
                else
                   argument = "float32ScaledToBitfield((float)" + argument;
            }
        }
        else
            argument = "float32ScaledToBitfield((float)" + argument;

        argument += ", " + getNumberString(encodedMin, encodedType.bits);
        argument += ", " + getNumberString(scaler, encodedType.bits);
        argument += ", " + std::to_string(encodedType.bits);
        argument += ")";
    }
    else if(isIntegerScaling())
    {
        if((encodedType.bits > 32) && support.longbitfield)
            argument = inMemoryType.toSigString() + "ScaledToLongBitfield(" + argument;
        else
            argument = inMemoryType.toSigString() + "ScaledToBitfield(" + argument;

        argument += ", " + std::to_string((int64_t)round(encodedMin));
        argument += ", " + std::to_string((int64_t)round(scaler));
        argument += ", " + std::to_string(encodedType.bits);
        argument += ")";
    }

    if(usesEncodeTempBitfield())
    {
        output += TAB_IN + "_pg_tempbitfield = (unsigned int)" + argument + ";\n";
        argument = "_pg_tempbitfield";
    }
    else if(usesEncodeTempLongBitfield())
    {
        output += TAB_IN + "_pg_templongbitfield = (uint64_t)" + argument + ";\n";
        argument = "_pg_templongbitfield";
    }

    if(bitfieldData.groupMember)
        output += ProtocolBitfield::getEncodeString(TAB_IN, argument, "_pg_bitfieldbytes", "_pg_bitfieldindex", bitfieldData.startingBitCount, encodedType.bits);
    else
        output += ProtocolBitfield::getEncodeString(TAB_IN, argument, "_pg_data", "_pg_byteindex", bitfieldData.startingBitCount, encodedType.bits);

    // Keep track of the total bits
    *bitcount += encodedType.bits;

    if(bitfieldData.lastBitfield)
    {
        if((bitfieldData.groupMember) && (bitfieldData.groupBits > 0))
        {
            // Number of bytes needed for all the bits
            int num = ((bitfieldData.groupBits+7)/8);

            output += "\n";

            output += TAB_IN + "// Encode the entire group of bits in one shot\n";

            if(support.bigendian)
                output += TAB_IN + "bytesToBeBytes(_pg_bitfieldbytes, _pg_data, &_pg_byteindex, " + std::to_string(num) + ");\n";
            else
                output += TAB_IN + "bytesToLeBytes(_pg_bitfieldbytes, _pg_data, &_pg_byteindex, " + std::to_string(num) + ");\n";

            output += TAB_IN + "_pg_bitfieldindex = 0;\n\n";

        }// if terminating a group
        else if((*bitcount) != 0)
        {
            // Increment our byte counter, 1 to 8 bits should result in 1 byte, 9 to 16 bits in 2 bytes, etc.
            int bytes = ((*bitcount)+7)/8;

            output += TAB_IN + "_pg_byteindex += " + std::to_string(bytes) + "; // close bit field\n\n";

        }// else if terminating a non-group

        // Reset bit counter
        *bitcount = 0;

    }// if this is a terminating bit field

    return output;

}// ProtocolField::getEncodeStringForBitfield


/*!
 * Get the next lines(s) of source coded needed to decode this bitfield field
 * \param bitcount points to the running count of bits in this string of
 *        bitfields, and will be updated by this fields encoded bit count.
 * \param isStructureMember should be true if the left hand side is a
 *        member of a user structure, else the left hand side is a pointer
 *        to the inMemoryType
 * \param defaultEnabled should be true to enable defaults for this decode
 * \return The string to add to the source file that decodes this field.
 */
std::string ProtocolField::getDecodeStringForBitfield(int* bitcount, bool isStructureMember, bool defaultEnabled) const
{
    std::string output;

    if(encodedType.isNull)
        return output;

    // If this field has a default value, or overrides a previous value
    if(defaultEnabled && (!defaultString.empty() || overridesPrevious))
    {
        int bytes;

        std::string lengthString;

        // How many bytes do we need? From 1 to 8 bits we need 1 byte, from
        // 9 to 15 we need 2 bytes, etc. However, some bits may already have
        // gone by
        if(bitfieldData.groupStart)
            bytes = (bitfieldData.groupBits+7)/8;
        else
            bytes = ((*bitcount) + encodedType.bits + 7)/8;

        lengthString = std::to_string(bytes);

        // Technically we only need to check the length if bitcount is zero
        // (i.e. a new byte is being decoded) or if this bitfield will cross
        // a byte boundary (again, requiring a new byte)
        if(((*bitcount) == 0) || (bytes > 1))
        {
            output += TAB_IN + "if(_pg_byteindex + " + lengthString + " > _pg_numbytes)\n";
            output += TAB_IN + TAB_IN + "return 1;\n";
            output += "\n";
        }
    }

    if(bitfieldData.groupStart)
    {
        int num = (bitfieldData.groupBits+7)/8;
        output += TAB_IN + "// Decode the entire group of bits in one shot\n";
        if(support.bigendian)
            output += TAB_IN + "bytesFromBeBytes(_pg_bitfieldbytes, _pg_data, &_pg_byteindex, " + std::to_string(num) + ");\n";
        else
            output += TAB_IN + "bytesFromLeBytes(_pg_bitfieldbytes, _pg_data, &_pg_byteindex, " + std::to_string(num) + ");\n";

        output += "\n";
    }

    if(!comment.empty())
        output += TAB_IN + "// " + comment + "\n";

    // Commenting indicating the range of the field
    if(!inMemoryType.isNull && !inMemoryType.isBool && (encodedType.bits > 1) && !inMemoryType.isEnum)
        output += TAB_IN + getRangeComment();

    std::string bitssource;
    std::string bitsindex;
    if(bitfieldData.groupMember)
    {
        bitssource = "_pg_bitfieldbytes";
        bitsindex = "_pg_bitfieldindex";
    }
    else
    {
        bitssource = "_pg_data";
        bitsindex = "_pg_byteindex";
    }

    // Handle the case where we just want to skip some bits
    if(inMemoryType.isNull && !checkConstant)
    {
        // Nothing to do in this case, it all gets handled when the bitfield terminates
    }
    else
    {        
        std::string argument;
        std::string cast;

        if(inMemoryType.isBool)
        {
            argument = getDecodeFieldAccess(isStructureMember);

            if(usesDecodeTempBitfield())
            {
                // This decodes the bitfield into the temporary variable
                output += ProtocolBitfield::getDecodeString(TAB_IN, "_pg_tempbitfield", cast, bitssource, bitsindex, *bitcount, encodedType.bits);

                // This tests the temporary variable and sets the boolean
                output += TAB_IN + argument + " = (_pg_tempbitfield) ? true : false;\n";
            }
            else if(usesDecodeTempLongBitfield())
            {
                // This decodes the bitfield into the temporary variable
                output += ProtocolBitfield::getDecodeString(TAB_IN, "_pg_templongbitfield", cast, bitssource, bitsindex, *bitcount, encodedType.bits);

                // This tests the temporary variable and sets the boolean
                output += TAB_IN + argument + " = (_pg_templongbitfield) ? true : false;\n";
            }
            else
            {
                // This tests the bitfield and sets the boolean
                output += TAB_IN + argument + " = (" + ProtocolBitfield::getInnerDecodeString(bitssource, bitsindex, *bitcount, encodedType.bits) + ") ? true : false;\n";
            }

        }// if in-memory is bool
        else
        {
            // How we are going to access the field
            if(usesDecodeTempBitfield())
                argument = "_pg_tempbitfield";
            else if(usesDecodeTempLongBitfield())
                argument = "_pg_templongbitfield";
            else
            {
                // If we are going to assign this bitfield directly to an enumeration,
                // and we do not have an intermediate temporary field, we need a cast
                if(inMemoryType.isEnum)
                    cast = "(" + typeName + ")";

                argument = getDecodeFieldAccess(isStructureMember);
            }

            // The argument in this case is a temporary if we are scaling, or we are inMemoryType.isNull. Otherwise this sets the actual in memory value
            output += ProtocolBitfield::getDecodeString(TAB_IN, argument, cast, bitssource, bitsindex, *bitcount, encodedType.bits);

            // Do the assignment from the temporary field
            if(!inMemoryType.isNull && (usesDecodeTempBitfield() || usesDecodeTempLongBitfield()))
            {
                output += TAB_IN + getDecodeFieldAccess(isStructureMember) + " = ";

                if(isFloatScaling())
                {
                    std::string cast;

                    if(!inMemoryType.isFloat)
                        cast = "(" + typeName +")";

                    if(encodedType.bits > 32)
                    {
                        if((inMemoryType.bits != 64) && support.float64)
                            cast = "(" + typeName +")";

                        if(support.longbitfield)
                        {
                            if(support.float64)
                                output += cast + "float64ScaledFromLongBitfield(" + argument;
                            else
                                output += cast + "float32ScaledFromLongBitfield(" + argument;
                        }
                        else
                        {
                            if(support.float64)
                                output += cast + "float64ScaledFromBitfield(" + argument;
                            else
                                output += cast + "float32ScaledFromBitfield(" + argument;
                        }
                    }
                    else
                        output += cast + "float32ScaledFromBitfield(" + argument;

                    output += ", " + getNumberString(encodedMin, encodedType.bits);
                    output += ", " + getNumberString(1.0, encodedType.bits) + "/" + getNumberString(scaler, encodedType.bits);
                    output += ");\n";

                }// if float scaled bitfield
                else if(isIntegerScaling())
                {
                    if(scaler != 1.0)
                    {
                        if((encodedType.bits > 32) && support.longbitfield)
                            output += inMemoryType.toSigString() + "ScaledFromLongBitfield(" + argument;
                        else
                            output += inMemoryType.toSigString() + "ScaledFromBitfield(" + argument;

                        output += ", " + std::to_string((int64_t)round(encodedMin)) + ", " + std::to_string((int64_t)round(scaler)) + ");\n";
                    }
                    else
                    {
                        // If the scaler is 1, add the min value directly - save the cost of the division
                        output += "(" + std::to_string((int64_t)round(encodedMin)) + " + " + argument + ");\n";
                    }

                }// If integer scaling is being done
                else
                {
                    // If we are going to assign directly to an enumeration we need a cast
                    if(inMemoryType.isEnum)
                        cast = "(" + typeName + ")";
                    else
                        cast.clear();

                    output += cast + argument + ";\n";

                }// else no scaling, some other reason for the temporary

            }// if a temporary value needs to get assigned to inMemory value

        }// else not bool

        if(checkConstant)
        {            
            std::string constantstring = getConstantString();

            // Verify the constant value
            output += TAB_IN + "// Decoded value must be " + constantstring + "\n";
            output += TAB_IN + "if(" + argument + " != " + constantstring + ")\n";
            output += TAB_IN + TAB_IN + "return " + getReturnCode(false) + ";\n";
        }

    }// else if we do something with the decoded value

    // Keep track of the number of bitfield bits that go by
    *bitcount += encodedType.bits;

    if(bitfieldData.lastBitfield)
    {
        if((bitfieldData.groupMember) && (bitfieldData.groupBits > 0))
        {
            output += TAB_IN + "_pg_bitfieldindex = 0;\n";

        }// if terminating a group
        else if((*bitcount) != 0)
        {
            // Increment our byte counter, 1 to 8 bits should result in 1 byte, 9 to 16 bits in 2 bytes, etc.
            int bytes = ((*bitcount)+7)/8;

            output += TAB_IN + "_pg_byteindex += " + std::to_string(bytes) + "; // close bit field\n\n";

        }// else if terminating a non-group

        output += "\n";

        // Reset bit counter
        *bitcount = 0;

    }// if this is a terminating bit field

    return output;

}// ProtocolField::getDecodeStringForBitfield


/*!
 * Get the source needed to close out a string of bitfields.
 * \param bitcount points to the running count of bits in this string of
 *        bitfields, and will be updated by this function.
 * \return The string to add to the source file that closes the bitfield.
 */
std::string ProtocolField::getCloseBitfieldString(int* bitcount) const
{
    std::string output;
    std::string spacing;

    if(*bitcount != 0)
    {
        // The number of bytes that the previous sequence of bit fields used up
        int length = *bitcount / 8;

        // Get the spacing right
        spacing += "    ";

        // If bitcount is not modulo 8, then the last byte was still in
        // progress, so increment past that
        if((*bitcount) % 8)
        {
            output += spacing + "bitcount = 0; _pg_byteindex++; // close bit field, go to next byte\n";
            length++;
        }
        else
        {
           output += spacing + "bitcount = 0; // close bit field, byte index already advanced\n";
        }

        output += "\n";

        // Reset bit counter
        *bitcount = 0;
    }

    return output;

}// ProtocolField::getCloseBitfieldString


/*!
 * Get the next lines of source needed to encode this string field.
 * \param isStructureMember should be true if the left hand side is a
 *        member of a user structure, else the left hand side is a pointer
 *        to the inMemoryType
 * \return The string to add to the source file to that encodes this field
 */
std::string ProtocolField::getEncodeStringForString(bool isStructureMember) const
{
    std::string output;

    std::string constantstring = getConstantString();

    if(encodedType.isNull)
        return output;

    if(!comment.empty())
        output += "    // " + comment + "\n";

    if(constantstring.empty())
        output += "    stringToBytes(" + getEncodeFieldAccess(isStructureMember) + ", _pg_data, &_pg_byteindex, " + array;
    else
        output += "    stringToBytes(" + constantstring + ", _pg_data, &_pg_byteindex, " + array;

    if(inMemoryType.isFixedString)
        output += ", 1);\n";
    else
        output += ", 0);\n";

    return output;

}// ProtocolField::getEncodeStringForString


/*!
 * Get the next lines of source needed to decode this string field
 * \param isStructureMember should be true if the left hand side is a
 *        member of a user structure, else the left hand side is a pointer
 *        to the inMemoryType
 * \param defaultEnabled should be true to handle defaults
 * \return The string to add to the source file to that decodes this field
 */
std::string ProtocolField::getDecodeStringForString(bool isStructureMember, bool defaultEnabled) const
{
    std::string output;
    std::string spacing = TAB_IN;

    if(encodedType.isNull)
        return output;

    if(!comment.empty())
        output += spacing + "// " + comment + "\n";

    // If this field has a default value, or overrides a previous value
    if(defaultEnabled && (!defaultString.empty() || overridesPrevious))
    {
        if(inMemoryType.isFixedString)
        {
            // If its a fixed string we must have the full monty
            output += spacing + "if(_pg_byteindex + " + array + " > _pg_numbytes)\n";
            output += spacing + TAB_IN + "return 1;\n";
            output += "\n";
            output += spacing + "stringFromBytes(" + getDecodeFieldAccess(isStructureMember) + ", _pg_data, &_pg_byteindex, " + array + ", 1);\n";
        }
        else
        {
            // Normal strings can be as short as a single character (empty strings with just a null)
            output += spacing + "if(_pg_byteindex + 1 > _pg_numbytes)\n";
            output += spacing + TAB_IN + "return 1;\n";
            output += "\n";

            // When pulling the bytes we have to control the maximum, it could be limited by the in memory space, or by the packet size
            output += spacing + "stringFromBytes(" + getDecodeFieldAccess(isStructureMember) + ", _pg_data, &_pg_byteindex, " + array + " < (_pg_numbytes - _pg_byteindex) ? " + array + " : (_pg_numbytes - _pg_byteindex), 0);\n";
        }
    }
    else
    {
        if(inMemoryType.isFixedString)
        {
            output += spacing + "stringFromBytes(" + getDecodeFieldAccess(isStructureMember) + ", _pg_data, &_pg_byteindex, " + array + ", 1);\n";
        }
        else
        {
            output += spacing + "stringFromBytes(" + getDecodeFieldAccess(isStructureMember) + ", _pg_data, &_pg_byteindex, " + array + ", 0);\n";
        }
    }

    if(checkConstant)
    {
        std::string constantstring = getConstantString();
        output += "\n";
        output += spacing + "// Decoded value must be " + constantstring + "\n";
        output += spacing + "if (strncmp(" + getDecodeFieldAccess(isStructureMember) + ", " + constantstring + ", " + array + ") != 0)\n";
        output += spacing + TAB_IN + "return " + getReturnCode(false) + ";\n";
    }

    return output;

}// ProtocolField::getDecodeStringForString


/*!
 * Return the string that is used to encode this structure.
 * \param isStructureMember is true if this encodable is accessed by structure pointer
 * \return the string to add to the source to encode this structure
 */
std::string ProtocolField::getEncodeStringForStructure(bool isStructureMember) const
{
    std::string output;
    std::string access;
    std::string spacing = TAB_IN;

    if(encodedType.isNull)
        return output;

    if(!comment.empty())
        output += spacing + "// " + comment + "\n";

    if(!dependsOn.empty())
    {
        output += spacing + "if(" + getEncodeFieldAccess(isStructureMember, dependsOn);

        if(!dependsOnValue.empty())
            output += " " + dependsOnCompare + " " + dependsOnValue;

        output += ")\n" + spacing + "{\n";
        spacing += TAB_IN;
    }

    // The array iteration code
    output += getEncodeArrayIterationCode(spacing, isStructureMember);

    // Spacing for arrays
    if(isArray())
    {
        spacing += TAB_IN;
        if(is2dArray())
            spacing += TAB_IN;
    }

    access = getEncodeFieldAccess(isStructureMember);

    if(support.language == ProtocolSupport::c_language)
        output += spacing + "encode" + typeName + "(_pg_data, &_pg_byteindex, " + access + ");\n";
    else
    {
        const ProtocolStructure* struc = parser->lookUpStructure(typeName);

        if((struc != nullptr) && (typeName != struc->getStructName()))
        {
            // We have an interesting case to handle here. This struct is
            // redefining another struct. In which case we need to call the
            // redefined encode function, and we do that by casting a pointer.
            // This is a downcast, which is only valid if there are no virtual
            // interfaces between here and the data we are going to touch.
            /// TODO: is there a better way? A very complex topic...
            if(isStructureMember || isArray())
                output += spacing + "(static_cast<const " + typeName + "*>(&" + access + "))->encode(_pg_data, &_pg_byteindex);\n";
            else
                output += spacing + "(static_cast<const " + typeName + "*>(" + access + "))->encode(_pg_data, &_pg_byteindex);\n";
        }
        else if(isStructureMember || isArray())
            output += spacing + access + ".encode(_pg_data, &_pg_byteindex);\n";
        else
            output += spacing + access + "->encode(_pg_data, &_pg_byteindex);\n";
    }

    if(!dependsOn.empty())
        output += TAB_IN + "}\n";

    return output;

}// ProtocolField::getEncodeStringForStructure


/*!
 * Get the next lines of source needed to decode this external structure field
 * \param isStructureMember should be true if the left hand side is a
 *        member of a user structure, else the left hand side is a pointer
 *        to the inMemoryType
 * \return The string to add to the source file to that decodes this field
 */
std::string ProtocolField::getDecodeStringForStructure(bool isStructureMember) const
{
    std::string output;
    std::string access;
    std::string spacing = TAB_IN;

    if(encodedType.isNull)
        return output;

    if(!comment.empty())
        output += "    // " + comment + "\n";

    if(!dependsOn.empty())
    {
        output += spacing + "if(" + getDecodeFieldAccess(isStructureMember, dependsOn);

        if(!dependsOnValue.empty())
            output += " " + dependsOnCompare + " " + dependsOnValue;

        output += ")\n" + spacing + "{\n";
        spacing += TAB_IN;
    }

    // The array iteration code
    output += getDecodeArrayIterationCode(spacing, isStructureMember);

    // Spacing for arrays
    if(isArray())
    {
        spacing += TAB_IN;
        if(is2dArray())
            spacing += TAB_IN;
    }

    access = getDecodeFieldAccess(isStructureMember);

    if(support.language == ProtocolSupport::c_language)
    {
        output += spacing + "if(decode" + typeName + "(_pg_data, &_pg_byteindex, " + access + ") == 0)\n";
        output += spacing + TAB_IN + "return 0;\n";
    }
    else
    {
        const ProtocolStructure* struc = parser->lookUpStructure(typeName);

        if((struc != nullptr) && (typeName != struc->getStructName()))
        {
            // We have an interesting case to handle here. This struct is
            // redefining another struct. In which case we need to call the
            // redefined encode function, and we do that by casting a pointer.
            // This is a downcast, which is only valid if there are no virtual
            // interfaces between here and the data we are going to touch.
            /// TODO: is there a better way? A very complex topic...
            if(isStructureMember || isArray())
                output += spacing + "if((static_cast<" + typeName + "*>(&" + access + "))->decode(_pg_data, &_pg_byteindex) == false)\n";
            else
                output += spacing + "if((static_cast<" + typeName + "*>(" + access + "))->decode(_pg_data, &_pg_byteindex) == false)\n";
        }
        else if(isStructureMember || isArray())
            output += spacing + "if(" + access + ".decode(_pg_data, &_pg_byteindex) == false)\n";
        else
            output += spacing + "if(" + access + "->decode(_pg_data, &_pg_byteindex) == false)\n";

        output += spacing + TAB_IN + "return false;\n";
    }

    if(!dependsOn.empty())
        output += TAB_IN + "}\n";

    return output;

}// ProtocolField::getDecodeStringForStructure


/*!
 * Look for a constantValue, in order of preference:
 * 1. constantValue
 * 2. If in memory type is null use "0"
 * \return the constant value
 */
std::string ProtocolField::getConstantString(void) const
{
    if (!constantString.empty())
    {
        if(encodedType.isString)
        {
            // constantValue is a string literal, so include the quotes. Except for
            // a special case. If constantValue ends in "()" then we assume its a
            // function or macro call
            if(endsWith(trimm(constantString), "()"))
                return constantString;
            else
                return "\"" + constantString + "\"";
        }
        else
            return constantString;
    }
    else if(inMemoryType.isNull)
    {
        if(encodedType.isString)
        {
            // A zero with quotes around it
            return "\"0\"";
        }
        else
            return "0";
    }
    else
        return std::string();

}// ProtocolField::getConstantString


/*!
 * Get the next lines(s) of source coded needed to encode this field, which
 * is not a bitfield or a string
 * \param isBigEndian should be true for big endian encoding.
 * \param isStructureMember should be true if the left hand side is a
 *        member of a user structure, else the left hand side is a pointer
 *        to the inMemoryType
 * \return The string to add to the source file that encodes this field.
 */
std::string ProtocolField::getEncodeStringForField(bool isBigEndian, bool isStructureMember) const
{
    std::string output;
    std::string endian;
    std::string lengthString;

    std::string constantstring = getConstantString();

    if(encodedType.isNull)
        return output;

    std::string spacing = TAB_IN;

    if(!comment.empty())
        output += spacing + "// " + comment + "\n";

    // Additional commenting to describe the scaling
    if(!inMemoryType.isNull && !inMemoryType.isBool && constantstring.empty())
        output += spacing + getRangeComment(support.limitonencode);

    int length = encodedType.bits / 8;
    lengthString = std::to_string(length);

    // Remember that we could be encoding an array
    if(isArray())
        lengthString += "*" + array;

    if(is2dArray())
        lengthString += "*" + array2d;

    // The endian string, which is empty for 1 byte, since
    // endian only applies to multi-byte fields
    if(length > 1)
    {
        if(isBigEndian)
            endian += "Be";
        else
            endian += "Le";
    }
    else
        endian = "";

    if(!dependsOn.empty())
    {
        output += spacing + "if(" + getEncodeFieldAccess(isStructureMember, dependsOn);

        if(!dependsOnValue.empty())
            output += " " + dependsOnCompare + " " + dependsOnValue;

        output += ")\n" + spacing + "{\n";
        spacing += TAB_IN;
    }

    std::string arrayspacing;
    std::string argument = getEncodeFieldAccess(isStructureMember);

    // The array iteration code
    output += getEncodeArrayIterationCode(spacing, isStructureMember);

    // Spacing for the array cases
    if(isArray())
    {
        arrayspacing += TAB_IN;
        if(is2dArray())
            arrayspacing += TAB_IN;
    }

    // Argument may need to be limited. Constant encode overrides the argument
    if(constantstring.empty())
        argument = getLimitedArgument(argument);
    else
        argument = constantstring;

    if(encodedType.isFloat)
    {
        // In this case we are encoding as a floating point. Typically we
        // would not scale here, but there are cases where scaling is
        // interesting.
        std::string scalestring;

        // Notice that encodedMax and encodedMin do not make sense since
        // the encoded type is float
        if(scaler != 1.0)
            scalestring = "*" + getNumberString(scaler, inMemoryType.bits);

        // Notice that we have to cast to the
        // input parameter type, since the user might (for example) have
        // the in-memory type as a double, but the encoded as a float
        std::string cast = "(" + encodedType.toTypeString() + ")";

        output += spacing + arrayspacing + "float" + std::to_string(encodedType.bits) + "To" + endian + "Bytes(" + cast + argument + scalestring + ", _pg_data, &_pg_byteindex";

        if((encodedType.bits == 16) || (encodedType.bits == 24))
            output += ", " + std::to_string(encodedType.sigbits);

        output += ");\n";

    }// If the encoded type is floating point
    else if(isFloatScaling() || isIntegerScaling())
    {
        std::string cast;

        output += spacing + arrayspacing;

        // The in-memory part of the scaling
        if(isFloatScaling())
        {
            // If we are floating scaling, we are going to use a float-encoding
            // function, since even an integer encoding function would still
            // have to cast to float to apply the scaler.
            if((inMemoryType.bits > 32) && support.float64)
            {
                output += "float64";

                // We may need a cast to float
                if(!inMemoryType.isFloat)
                    cast = "(double)";
            }
            else
            {
                output += "float32";

                // We may need a cast to float
                if((!inMemoryType.isFloat) || ((constantstring.find('.') != std::string::npos) && constantstring.back() != 'f'))
                    cast = "(float)";
            }

        }// if float scaling
        else
        {
            output += inMemoryType.toSigString(); // "int16" or "uint32" for example

        }// else not float scaling

        // More of the encode function name includuing number of bytes
        output += "ScaledTo" + std::to_string(length);

        // Signed or unsigned
        if(encodedType.isSigned)
            output += "Signed";
        else
            output += "Unsigned";

        // More of the encode call signature
        output += endian + "Bytes(" + cast + argument + ", _pg_data, &_pg_byteindex";

        if(isFloatScaling())
        {
            // Signature changes for signed versus unsigned
            if(!encodedType.isSigned)
                output += ", " + getNumberString(encodedMin, inMemoryType.bits);

            output +=  ", " + getNumberString(scaler, inMemoryType.bits);
        }
        else
        {
            // Signature changes for signed versus unsigned
            if(!encodedType.isSigned)
                output += ", " + std::to_string((int64_t)round(encodedMin));

            output +=  ", " + std::to_string((int64_t)round(scaler));
        }

        output += ");\n";

    }// else if scaling (floating or integer)
    else
    {
        std::string function;

        if(encodedType.isSigned)
        {
            // "int32ToBeBytes(" for example
            function = "int" + std::to_string(encodedType.bits) + "To" + endian + "Bytes(";
        }
        else
        {
            // "uint32ToBeBytes(" for example
            function = "uint" + std::to_string(encodedType.bits) + "To" + endian + "Bytes(";
        }

        // Cast the constant string, just in case
        if(!constantstring.empty())
        {
            // "uint32ToBeBytes((uint32_t)(CONSTANT)" for example
            function += "(" + encodedType.toTypeString() + ")(" + constantstring + ")";
        }
        else
        {
            if(inMemoryType.isBool)
            {
                function += "(" + argument + " == true) ? 1 : 0";
            }
            else if(inMemoryType.isFloat || (inMemoryType.bits > encodedType.bits) || (inMemoryType.isSigned != encodedType.isSigned))
            {
                // Add a cast in case the encoded type is different from the in memory type
                function += "(" + encodedType.toTypeString() + ")(" + argument + ")";
            }
            else
                function += argument;

        }// else if not constant

        // This is the termination of the function and the line
        function += ", _pg_data, &_pg_byteindex);\n";

        output += spacing + arrayspacing + function;

    }// else if not float scaled

    if(!dependsOn.empty())
        output += TAB_IN + "}\n";

    return output;

}// ProtocolField::getEncodeStringForField


/*!
 * Check to see if we should be doing floating point scaling on this field.
 * This means the encode operation is going to call a function like
 * `float32To...()`; and the decode function is going to call a function like
 * `float32From...()`.
 * \return true if scaling should be based on floating point operations
 */
bool ProtocolField::isFloatScaling(void) const
{
    // If the encoding is floating point then we are not scaling
    if(encodedType.isFloat)
        return false;

    if(inMemoryType.isFloat)
    {
        // If the user input a scaler string, then they want float scaling
        if(!scalerString.empty() || (scaler != 1.0))
           return true;

        // The min value only matters if we are signed
        if(!encodedType.isSigned && (encodedMin != 0.0))
            return true;
    }
    else
    {
        // In this case the inMemoryType is not a float, but we might still
        // want float scaling if the scaling terms are not integer
        if((scaler != 1.0) && (ceil(scaler) != floor(scaler)))
            return true;

        // The min value only matters if we are signed
        if(!encodedType.isSigned && (encodedMin != 0.0) && (ceil(encodedMin) != floor(encodedMin)))
            return true;
    }

    return false;

}// ProtocolField::isFloatScaling


/*!
 * Check to see if we should be doing integer scaling on this field. This
 * means encode and decode functions will inline scaling operations using integers.
 * \return true if scaling should be based on integer operations
 */
bool ProtocolField::isIntegerScaling(void) const
{
    // If the encoding is floating point then we are not scaling
    if(encodedType.isFloat)
        return false;

    if(!inMemoryType.isFloat)
    {
        if((scaler != 1.0) && (ceil(scaler) == floor(scaler)))
            return true;

        // The min value only matters if we are signed
        if(!encodedType.isSigned && (encodedMin != 0.0) && (ceil(encodedMin) == floor(encodedMin)))
            return true;
    }

    return false;

}// ProtocolField::isIntegerScaling


/*!
 * Get the next lines(s) of source coded needed to decode this field, which
 * is not a bitfield or a string
 * \param isBigEndian should be true if the protocol uses big endian ordering.
 * \param isStructureMember should be true if the left hand side is a
 *        member of a user structure, else the left hand side is a pointer
 *        to the inMemoryType
 * \param defaultEnabled should be true to enable default handling
 * \return The string to add to the source file that decodes this field.
 */
std::string ProtocolField::getDecodeStringForField(bool isBigEndian, bool isStructureMember, bool defaultEnabled) const
{
    std::string output;
    std::string endian;
    std::string spacing = TAB_IN;
    std::string arrayspacing;
    std::string argument;

    std::string constantstring = getConstantString();

    if(encodedType.isNull)
        return output;

    argument = getDecodeFieldAccess(isStructureMember);

    int length = encodedType.bits / 8;

    // The endian string, which is empty for 1 byte, since
    // endian only applies to multi-byte fields
    if(length > 1)
    {
        if(isBigEndian)
            endian += "Be";
        else
            endian += "Le";
    }
    else
        endian = "";

    // What is the length in bytes of this field, remember that we could be encoding an array
    std::string lengthString = std::to_string(length);

    if(isArray())
    {
        if(variableArray.empty())
            lengthString += "*" + array;
        else
            lengthString += "*" + getDecodeFieldAccess(isStructureMember, variableArray);

        if(is2dArray())
        {
            if(variable2dArray.empty())
                lengthString += "*" + array2d;
            else
                lengthString += "*" + getDecodeFieldAccess(isStructureMember, variable2dArray);
        }
    }

    if(!dependsOn.empty())
    {
        output += spacing + "if(" + getDecodeFieldAccess(isStructureMember, dependsOn);

        if(!dependsOnValue.empty())
            output += " " + dependsOnCompare + " " + dependsOnValue;

        output += ")\n" + spacing + "{\n";
        spacing += TAB_IN;
    }

    // If this field has a default value, or overrides a previous value
    if(defaultEnabled && (!defaultString.empty() || overridesPrevious))
    {
        output += spacing + "if(_pg_byteindex + " + lengthString + " > _pg_numbytes)\n";
        output += spacing + TAB_IN + "return " + getReturnCode(true) + ";\n";
        output += "\n";
    }

    if(!comment.empty())
        output += spacing + "// " + comment + "\n";

    // Commenting indicating the range of the field
    if(!inMemoryType.isNull && !inMemoryType.isBool)
        output += spacing + getRangeComment();

    if(!inMemoryType.isNull && checkConstant)
        output += spacing + "// Decoded value must be " + constantstring + "\n";

    if(inMemoryType.isNull)
    {
        if(checkConstant && array.empty())
        {
            output += spacing + "if (";

            if(encodedType.isFloat)
            {
                if(encodedType.bits == 16)
                    output += "float16From" + endian + "Bytes(_pg_data, &_pg_byteindex, " + std::to_string(encodedType.sigbits) + ")";
                else if(encodedType.bits == 24)
                    output += "float24From" + endian + "Bytes(_pg_data, &_pg_byteindex, " + std::to_string(encodedType.sigbits) + ")";
                else if((inMemoryType.bits > 32) && support.float64)
                    output += "float64From" + endian + "Bytes(_pg_data, &_pg_byteindex)";
                else
                    output += "float32From" + endian + "Bytes(_pg_data, &_pg_byteindex)";
            }
            else
            {
                if(encodedType.isSigned)
                    output += "int";
                else
                    output += "uint";

                output += std::to_string(encodedType.bits) + "From" + endian + "Bytes(_pg_data, &_pg_byteindex)";
            }

            output += " != (" + encodedType.toTypeString() + ") " + constantstring + ")\n";
            output += spacing + TAB_IN + "return " + getReturnCode(false) + ";\n";

        }// If constant value must be checked
        else
        {
            // Skip over reserved space
            if(comment.empty())
                output += spacing + "// Skip over reserved space\n";

            output += spacing + "_pg_byteindex += " + lengthString + ";\n";

        }// else constant value is not checked

    }// If nothing in-memory
    else
    {
        output += getDecodeArrayIterationCode(spacing, isStructureMember);

        // Array spacing
        if(isArray())
        {
            arrayspacing += TAB_IN;
            if(is2dArray())
                arrayspacing += TAB_IN;
         }

        if(encodedType.isFloat)
        {
            // In this case we are encoding as a floating point. Typically we
            // would not scale here, but there are cases where scaling is
            // interesting.
            std::string scalestring;

            // Notice that encodedMax and encodedMin do not make sense since
            // the encoded type is float
            if(scaler != 1.0)
                scalestring = "(" + getNumberString(1.0, inMemoryType.bits) + "/" + getNumberString(scaler, inMemoryType.bits) + ")*" ;

            output += spacing + arrayspacing + argument + " = " + scalestring + "float" + std::to_string(encodedType.bits) + "From" + endian + "Bytes(_pg_data, &_pg_byteindex";

            if((encodedType.bits == 16) || (encodedType.bits == 24))
                output += ", " + std::to_string(encodedType.sigbits);

            output += ");\n";

        }// if float
        else if(isFloatScaling())
        {
            output += spacing + arrayspacing + argument + " = ";

            // The cast if the in memory type is not floating
            if(!inMemoryType.isFloat)
                output += "(" + inMemoryType.toTypeString() + ")";

            if((inMemoryType.bits > 32) && support.float64)
                output += "float64";
            else
                output += "float32";

            // The scaling decode function
            output += "ScaledFrom" + std::to_string(length);

            // Signed or unsigned
            if(encodedType.isSigned)
                output += "Signed";
            else
                output += "Unsigned";

            output += endian + "Bytes(_pg_data, &_pg_byteindex";

            // Signature changes for signed versus unsigned
            if(!encodedType.isSigned)
                output += ", " + getNumberString(encodedMin, inMemoryType.bits);

            // Notice how the scaling value is the inverse for the decode function
            output += ", " + getNumberString(1.0, inMemoryType.bits) + "/" + getNumberString(scaler, inMemoryType.bits);

            output += ");\n";

        }// if float scaling to integer
        else if(isIntegerScaling())
        {
            output += spacing + arrayspacing + argument + " = ";

            // If the scaler is 1, add the minimum value in directly - save the cost of division
            if(scaler == 1.0)
            {
                // Might need a cast
                if((inMemoryType.isSigned != encodedType.isSigned) || (inMemoryType.bits < encodedType.bits))
                    output += "(" + inMemoryType.toTypeString() + ")";

                // "int32FromBeBytes(data, &_pg_byteindex)" for example
                output += "(" + encodedType.toSigString() + "From" + endian + "Bytes(_pg_data, &_pg_byteindex) + " + std::to_string((int64_t)round(encodedMin)) + ");\n";
            }
            else
            {
                // "uint32" or "int16" for example
                output += inMemoryType.toSigString();

                // Scaled from a number of bytes
                output += "ScaledFrom" + std::to_string(length);

                // Signed or unsigned
                if(encodedType.isSigned)
                    output += "Signed";
                else
                    output += "Unsigned";

                // In an endian order
                output += endian + "Bytes(_pg_data, &_pg_byteindex";

                // Signature changes for signed versus unsigned
                if(!encodedType.isSigned)
                    output += ", " + std::to_string((int64_t)round(encodedMin));

                output += ", " + std::to_string((int64_t)round(scaler));

                output += ");\n";
            }

        }// else if integer scaling to integer
        else
        {
            std::string function;

            if(encodedType.isSigned)
                function = "int";
            else
                function = "uint";

            // "int32FromBeBytes(data, &_pg_byteindex)" for example
            function += std::to_string(encodedType.bits) + "From" + endian + "Bytes(_pg_data, &_pg_byteindex)";

            if(inMemoryType.isBool)
            {
                function = "(" + function + ") ? true : false";
            }
            else if(inMemoryType.isFloat || (inMemoryType.bits != encodedType.bits) || inMemoryType.isEnum)
            {
                // Add a cast in case the encoded type is different from the in memory type
                // "int32ToBeBytes((int32_t)((_pg_user->value - min)*scale)" for example
                function = "(" + typeName + ")" + function;
            }

            output += spacing + arrayspacing + argument + " = " + function + ";\n";

        }// else not floating point scaled

    }// else not null in-memory

    // Handle the check constant case, the null case was handled above
    if(!inMemoryType.isNull && checkConstant)
    {
        output += spacing + arrayspacing + "if (" + argument + " != " + constantstring + ")\n";
        output += spacing + arrayspacing + TAB_IN + "return " + getReturnCode(false) + ";\n";
    }

    // Close the depends on block
    if(!dependsOn.empty())
        output += TAB_IN + "}\n";

    return output;

}// ProtocolField::getDecodeStringForField


/*!
 * Get a properly formatted number string for a floating point number.
 * \param number is the number to turn into a string.
 * \param bits is the number of bits in memory for this string. 32 or
 *        less will prompt a 'f' suffix on the string.
 * \return the string.
 */
std::string ProtocolField::getNumberString(double number, int bits) const
{
    std::string string;
    std::stringstream stream;

    // We'll use scientific notation for numbers with more than 9 digits
    if(fabs(number) > 999999999.0)
        stream << std::scientific;

    // Note DBL_DECIMAL_DIG is 17 in IEE-754. This results in unsightly
    // rounding, for example 65.534999999999997 instead of 65.535. Hence
    // use one less (which is what we used to have).
    if((bits <= 32) || !support.float64)
        stream << std::setprecision(std::numeric_limits<float>::max_digits10);
    else
        stream << std::setprecision(std::numeric_limits<double>::max_digits10 - 1);

    // Now put the number in the stream
    stream << number;

    // And get a string for it
    string = stream.str();

    // Floating point constants in code should have decimal places
    if(string.find('.') == std::string::npos)
        string += ".0";

    // Floating point constant
    if((bits <= 32) || !support.float64)
        string += "f";

    return string;

}// ProtocolField::getNumberString


/*!
 * Get a properly formatted number string for a floating point number. If the
 * number is one of these multiples of pi (-2, -1, -.5, .5, 1, 2), then return
 * a string that includes the html pi token.
 * \param number is the number to turn into a string
 * \return the string.
 */
std::string ProtocolField::getDisplayNumberString(double number)
{
    if(number == -2*3.14159265358979323846)
    {
        return "-2&pi;";
    }
    else if(number == -3.14159265358979323846)
    {
        return "-&pi;";
    }
    else if(number == -0.5*3.14159265358979323846)
    {
        return "-&pi;/2";
    }
    else if(number == 0.5*3.14159265358979323846)
    {
        return "&pi;/2";
    }
    else if(number == 3.14159265358979323846)
    {
        return "&pi;";
    }
    else if(number == 2*3.14159265358979323846)
    {
        return "2&pi;";
    }
    else
    {
        std::string string;
        std::stringstream stream;

        // We'll use scientific notation for numbers with more than 9 digits
        if(fabs(number) > 999999999.0)
            stream << std::scientific;

        // Note DBL_DECIMAL_DIG is 17 in IEE-754. This results in unsightly
        // rounding, for example 65.534999999999997 instead of 65.535. Hence
        // use one less (which is what we used to have).
        stream << std::setprecision(std::numeric_limits<double>::max_digits10 - 1);

        // Now put the number in the stream
        stream << number;

        // And get a string for it
        string = stream.str();

        // Floating point constants in code should have decimal places
        if(string.find('.') == std::string::npos)
            string += ".0";

        return string;
    }
}