Skip to content

Latest commit

 

History

History
206 lines (140 loc) · 10.5 KB

changelog.md

File metadata and controls

206 lines (140 loc) · 10.5 KB
Raw

Change Log

Version 2

This is a major new version with many new features added. We have been very careful to preserve backwards compatibility of the outward-facing API of the library while adding new features. The biggest change (and the reason for the major version number) is that we are now fully compatible with python versions 2.7, 3.6 and 3.7. All typical applications of version 1.2 should be compatible with version 2. However, we do warn that we broke internal backwards compatibility in a number of locations in order to accommodate python 3. If you previously wrote custom plugins for version 1.2, we cannot guarantee that they will continue to work in version 2.

Version 2.4

Version 2.4.1

  • Implemented the option to specify a tuple of multiple possible answers in an 'expect' keyword when supplying specific messages and grades.

Version 2.4.0

  • Refactored loading of scipy to only occur when required. This should make the library load significantly faster, which will resolve a lot of codejail timeout errors (we hope!).

Version 2.3

Version 2.3.3

  • Refactored sampling to allow sibling variables to be used in DependentSamplers.

Version 2.3.2

  • Fixed a bug where long integers passed in as a variable (via DiscreteSet) or user-defined constant would cause an error in evaluation (such numbers are now cast as floats).

Version 2.3.1

  • Added a feature to math graders to remove default constants such as pi or i.
  • Updated math parser to treat emdash unicode characters as minus signs (thanks Derek Rhodes!).
  • Updated pytest version to last version supporting python 2.

Version 2.3.0

  • Updated unit tests to use python and library versions consistent with edx.org.
  • Updated documentation to reflect edX usage of python 3.
  • Added python version to debug output.
  • Fixed a bug in LinearComparer that caused a crash when a student provided a constant answer.
  • Fixed a bug in debug output for math graders that caused output not to show.

Version 2.2

Version 2.2.0

  • Rewrote IntegralGrader and FormulaGrader to leverage the same underlying functions, which allows all of the options in FormulaGrader to also apply to IntegralGrader.
  • Constructed SumGrader, which operates like IntegralGrader, but grades sums instead of integrals.

Version 2.1

Version 2.1.1

  • Fixed a bug in LinearComparer when a student's answer was constant.

Version 2.1.0

  • Added a new grader IntervalGrader, which grades expressions for mathematical intervals.
  • Added allow_inf option to FormulaGrader and NumericalGrader, which allow for infty (infinity) to appear in expressions.
  • SingleListGrader can now be provided answers as a single string. It can also infer answers from the expect keyword of a customresponse tag.

Version 2.0

Version 2.0.2

  • Fix a bug with custom AsciiMath preprocessor where lonely square roots (e.g., x + sqrt + y) rendered badly.

Version 2.0.1

  • Fixed a bug in StringGrader regex validation that only required the start of student input to match the pattern.

Version 2.0.0

Feature updates:

  • Graders can now be constructed in-line in a customresponse, rather than in a python code block.
  • Credit awarded can now be dependent upon the attempt number.
  • Introduced a new plug-in to set course-wide defaults that differ from library defaults.
  • StringGrader received a major overhaul, and can now require minimum character or word lengths, and also pattern match to regular expressions. Further options to treat whitespace were introduced.
  • The structure of SingleListGrader answers now uses and extends that used by ItemGrader answers, leading to more general ways of expressing answers and providing feedback to students.
  • Answers to SingleListGrader problems can now be inferred from the expect keyword.
  • SingleListGrader can now warn students that they have missing entries.
  • Partial credit can now be turned off in ListGrader problems.
  • Partial credit can now be awarded in matrix entry problems, with messages explaining which entries are correct.
  • Implemented LinearComparer, which can assign partial/full credit to formula responses that are linearly related to answers.
  • Comparers for math problems can now be set much more straightforwardly.
  • floor, ceil, min and max functions were added to the math library.
  • In math problems, introduced an option to apply a transforming function to the answer and student input before comparing.
  • Completely overhauled matrix sampling to allow most typical matrix types to be sampled.
  • Introduced instructor variables for math problems, which authors can use in constructing the problem, but students may not use in their responses.
  • DependentSampler now has access to constants and user functions, and no longer needs the depends key.
  • DiscreteSet will now work with arrays.
  • All comparers are now imported when using from mitxgraders import *.

Under the hood:

  • Ensured that the entire library (including tests) is compatible with python 2.7 and python 3.6/3.7, in preparation for edX's upcoming transition to python 3. (The change to having internal unicode literals is the biggest internal backward-compatability-breaking issue from version 1.2.)
  • Introduced CorrelatedComparers, which compare multiple samples at once.
  • IntegralGrader has been promoted from a plug-in to a core component of the library.
  • SpecifyDomain has been expanded to allow for functions with any number of arguments.
  • Javascript preprocessor was tidied up, and further options for customization were included.
  • All documentation examples are now run as doctests.
  • Upgraded voluptuous to version 0.11.5.
  • Upgraded testing infrastructure.

Bug fixes:

  • Fixed a bug in assigning partial credit in ListGraders when multiple lists of answers were included.

Version 1

Version 1.2

Version 1.2.3

  • Added new custom comparers vector_span_comparer and vector_phase_comparer.
  • Updated AsciiMath preprocessor to handle lonely mover (math-over) entries.

Version 1.2.2

  • Added accept_any and accept_nonempty options to StringGrader.
  • If only a single text input is present, the answers key is inferred from the expect attribute of the customresponse tag (does not work for SingleListInput however).
  • Updated AsciiMath preprocessor to handle variable names and user-defined functions properly.
  • Various small bug fixes.

Version 1.2.1

  • Added arctan2(x, y) function that returns angle between +x axis and the point (x, y); available by default for students to use in FormulaGrader problems.
  • Rewrote the expression parser and evaluator. Among other things, new parser provides better error messages when explicit multiplication is forgotten in expressions like '5x + 3'. Also validates curly brace balancing, which may be used in tensor variable names.

Version 1.2.0

This version includes a number of new features and documentation updates.

  • A new documentation website
  • Math parser now supports multivariable functions and array input (vector, matrix, etc).
  • Many improvements to our mathjax preprocessor
  • Improvements to balanced bracket validator.
  • Added new class MatrixGrader (see MatrixGrader documentation) along with supporting sampling classes RealMatrices and RealVectors
  • When FormulaGrader (and its subclasses) are used inside an ordered ListGrader, authors can now grade multiple student inputs in comparison to each other by specifying answers in terms of sibling variables
  • FormulaGrader (and its subclasses) now support comparer functions that can be used to grade student input more flexibly. For example, rather than checking checking that the student input and author input are equal, check that they are equal modulo a certain number. Built-in comparers:
    • equality_comparer
    • congruence_comparer
    • between_comparer
    • eigenvector_comparer

Version 1.1

Version 1.1.2

  • This version includes an internal change to the way that errors are handled during check.
    • If you only use builtin graders (FormulaGrader, ListGrader...) or public plugins (IntegralGrader) you should not notice any difference.
    • If you have previously written your own grading class, this change could affect what errors messages are displayed to students. In particular, only exceptions inheriting from MITxError will display their messages to students; other errors will be replaced with a generic error message.

Version 1.1.1

  • Added AsciiMath renderer definitions
  • We now check for naming collisions in your configuration
  • Cleaned up voluptuous incorporation
  • Extend domain of factorial function to all complex, except negative integers
  • Removed .pyc files from the zip file
  • Minor bug fixes

Version 1.1.0

  • Added numbered variables to FormulaGrader

  • Removed case-insensitive comparisons from FormulaGrader and IntegralGrader.

    !!! warning This is a departure from edX and is a breaking change for authors who used case-insensitive FormulaGraders. However:

      - Case-sensitive has always been the default for FormulaGrader and we are not aware of authors using case-insensitive FormulaGraders.
  - Pedagogically, we believe that students should think of `M` and `m` are different variables.
  - Removing case-insensitive comparison fixes a number of ambiguous situations.

Version 1.0

Version 1.0.5

  • Improved debugging information for FormulaGrader
  • FormulaGrader and IntegralGrader perform whitelist, blacklist, and forbidden_string checks after determining answer correctness. Incorrect answers using forbidden strings / functions are now marked incorrect, while correct answers using forbidden strings / functions raise errors.
  • Minor improvements to existing unit tests

Version 1.0.4

  • Authors can now specify a custom comparer function for FormulaGrader
  • IntegralGrader now handles complex integrands, and gives meaningful error messages for complex limits.
  • Miscellaneous bug fixes for tensor variable name parsing

Version 1.0.3

  • Added tensor variable names

Version 1.0.2

  • Added error messages for overflow, division-by-zero, and out-of-domain errors in formulas
  • Added tests to reach 100% coverage
  • Removed redundant code
  • Fixed some bugs in unused code

Version 1.0.1

  • Added DependentSampler
  • Fixed issue with zip file tests
  • Added doctests to test suite
  • Fixed bug in FormulaGrader when given an empty string

Version 1.0.0

  • First release!