[FormRecognizer] Preview 1: detect content type when not provided

sdk/formrecognizer/Azure.AI.FormRecognizer/src/FormRecognizerClient.cs

@@ -2,7 +2,6 @@
 // Licensed under the MIT License.
 
 using System;
-using System.Collections.Generic;
 using System.IO;
 using System.Threading;
 using System.Threading.Tasks;
@@ -60,15 +59,11 @@ protected FormRecognizerClient()
         [ForwardsClientCalls]
         public virtual RecognizeContentOperation StartRecognizeContent(Stream formFileStream, RecognizeOptions recognizeOptions = default, CancellationToken cancellationToken = default)
         {
-            // TODO: automate content-type detection
-            // https://github.com/Azure/azure-sdk-for-net/issues/10329
-            ResponseWithHeaders<ServiceAnalyzeLayoutAsyncHeaders> response =  ServiceClient.RestClient.AnalyzeLayoutAsync(ContentType.Pdf, formFileStream, cancellationToken);
-            //Response response = ServiceClient.RestClient.AnalyzeLayoutAsync(ContentType.Pdf, formFileStream, cancellationToken);
-
-            // TODO: throw Exception if header is not present.
-            //response.Headers.TryGetValue("Operation-Location", out string operationLocation);
-            string operationLocation = response.Headers.OperationLocation;
-            return new RecognizeContentOperation(ServiceClient, operationLocation);
+            recognizeOptions ??= new RecognizeOptions();
+            ContentType contentType = recognizeOptions.ContentType ?? DetectContentType(formFileStream, nameof(formFileStream));
+
+            ResponseWithHeaders<ServiceAnalyzeLayoutAsyncHeaders> response =  ServiceClient.RestClient.AnalyzeLayoutAsync(contentType, formFileStream, cancellationToken);
+            return new RecognizeContentOperation(ServiceClient, response.Headers.OperationLocation);
         }
 
         /// <summary>
@@ -82,15 +77,11 @@ public virtual RecognizeContentOperation StartRecognizeContent(Stream formFileSt
         [ForwardsClientCalls]
         public virtual async Task<RecognizeContentOperation> StartRecognizeContentAsync(Stream formFileStream, RecognizeOptions recognizeOptions = default, CancellationToken cancellationToken = default)
         {
-            // TODO: automate content-type detection
-            // https://github.com/Azure/azure-sdk-for-net/issues/10329
-            ResponseWithHeaders<ServiceAnalyzeLayoutAsyncHeaders> response = await ServiceClient.RestClient.AnalyzeLayoutAsyncAsync(ContentType.Pdf, formFileStream, cancellationToken).ConfigureAwait(false);
-            //Response response = await ServiceClient.RestClient.AnalyzeLayoutAsyncAsync(ContentType.Pdf, formFileStream, cancellationToken).ConfigureAwait(false);
-
-            // TODO: throw Exception if header is not present.
-            //response.Headers.TryGetValue("Operation-Location", out string operationLocation);
-            string operationLocation = response.Headers.OperationLocation;
-            return new RecognizeContentOperation(ServiceClient, operationLocation);
+            recognizeOptions ??= new RecognizeOptions();
+            ContentType contentType = recognizeOptions.ContentType ?? DetectContentType(formFileStream, nameof(formFileStream));
+
+            ResponseWithHeaders<ServiceAnalyzeLayoutAsyncHeaders> response = await ServiceClient.RestClient.AnalyzeLayoutAsyncAsync(contentType, formFileStream, cancellationToken).ConfigureAwait(false);
+            return new RecognizeContentOperation(ServiceClient, response.Headers.OperationLocation);
         }
 
         /// <summary>
@@ -106,12 +97,7 @@ public virtual RecognizeContentOperation StartRecognizeContentFromUri(Uri formFi
         {
             SourcePath_internal sourcePath = new SourcePath_internal(formFileUri.ToString());
             ResponseWithHeaders<ServiceAnalyzeLayoutAsyncHeaders> response = ServiceClient.RestClient.AnalyzeLayoutAsync(sourcePath, cancellationToken);
-            //Response response = ServiceClient.RestClient.AnalyzeLayoutAsync(sourcePath, cancellationToken);
-
-            // TODO: throw Exception if header is not present.
-            //response.Headers.TryGetValue("Operation-Location", out string operationLocation);
-            string operationLocation = response.Headers.OperationLocation;
-            return new RecognizeContentOperation(ServiceClient, operationLocation);
+            return new RecognizeContentOperation(ServiceClient, response.Headers.OperationLocation);
         }
 
         /// <summary>
@@ -127,12 +113,7 @@ public virtual async Task<RecognizeContentOperation> StartRecognizeContentFromUr
         {
             SourcePath_internal sourcePath = new SourcePath_internal(formFileUri.ToString());
             ResponseWithHeaders<ServiceAnalyzeLayoutAsyncHeaders> response = await ServiceClient.RestClient.AnalyzeLayoutAsyncAsync(sourcePath, cancellationToken).ConfigureAwait(false);
-            //Response response = await ServiceClient.RestClient.AnalyzeLayoutAsyncAsync(sourcePath, cancellationToken).ConfigureAwait(false);
-
-            // TODO: throw Exception if header is not present.
-            //response.Headers.TryGetValue("Operation-Location", out string operationLocation);
-            string operationLocation = response.Headers.OperationLocation;
-            return new RecognizeContentOperation(ServiceClient, operationLocation);
+            return new RecognizeContentOperation(ServiceClient, response.Headers.OperationLocation);
         }
 
         #endregion
@@ -144,18 +125,15 @@ public virtual async Task<RecognizeContentOperation> StartRecognizeContentFromUr
         /// </summary>
         /// <param name="receiptFileStream">The stream containing the one or more receipts to recognize values from.</param>
         /// <param name="receiptLocale"></param>>
-        /// <param name="contentType"></param>
         /// <param name="recognizeOptions"></param>
         /// <param name="cancellationToken">A <see cref="CancellationToken"/> controlling the request lifetime.</param>
         /// <returns>A <see cref="RecognizeReceiptsOperation"/> to wait on this long-running operation.  Its <see cref="RecognizeReceiptsOperation"/>.Value upon successful
         /// completion will contain the extracted receipt.</returns>
         [ForwardsClientCalls]
-        public virtual async Task<RecognizeReceiptsOperation> StartRecognizeReceiptsAsync(Stream receiptFileStream, ContentType contentType, string receiptLocale = "en-US", RecognizeOptions recognizeOptions = default, CancellationToken cancellationToken = default)
+        public virtual async Task<RecognizeReceiptsOperation> StartRecognizeReceiptsAsync(Stream receiptFileStream, string receiptLocale = "en-US", RecognizeOptions recognizeOptions = default, CancellationToken cancellationToken = default)
         {
-            // TODO: automate content-type detection
-            // https://github.com/Azure/azure-sdk-for-net/issues/10329
-
             recognizeOptions ??= new RecognizeOptions();
+            ContentType contentType = recognizeOptions.ContentType ?? DetectContentType(receiptFileStream, nameof(receiptFileStream));
 
             ResponseWithHeaders<ServiceAnalyzeReceiptAsyncHeaders> response = await ServiceClient.RestClient.AnalyzeReceiptAsyncAsync(contentType, receiptFileStream, includeTextDetails: recognizeOptions.IncludeTextContent, cancellationToken).ConfigureAwait(false);
             return new RecognizeReceiptsOperation(ServiceClient, response.Headers.OperationLocation);
@@ -166,18 +144,15 @@ public virtual async Task<RecognizeReceiptsOperation> StartRecognizeReceiptsAsyn
         /// </summary>
         /// <param name="receiptFileStream">The stream containing the one or more receipts to recognize values from.</param>
         /// <param name="receiptLocale"></param>
-        /// <param name="contentType"></param>
         /// <param name="recognizeOptions">Whether or not to include raw page recognition in addition to layout elements.</param>
         /// <param name="cancellationToken">A <see cref="CancellationToken"/> controlling the request lifetime.</param>
         /// <returns>A <see cref="RecognizeReceiptsOperation"/> to wait on this long-running operation.  Its <see cref="RecognizeReceiptsOperation"/>.Value upon successful
         /// completion will contain the extracted receipt.</returns>
         [ForwardsClientCalls]
-        public virtual RecognizeReceiptsOperation StartRecognizeReceipts(Stream receiptFileStream, ContentType contentType, string receiptLocale = "en-US", RecognizeOptions recognizeOptions = default, CancellationToken cancellationToken = default)
+        public virtual RecognizeReceiptsOperation StartRecognizeReceipts(Stream receiptFileStream, string receiptLocale = "en-US", RecognizeOptions recognizeOptions = default, CancellationToken cancellationToken = default)
         {
-            // TODO: automate content-type detection
-            // https://github.com/Azure/azure-sdk-for-net/issues/10329
-
             recognizeOptions ??= new RecognizeOptions();
+            ContentType contentType = recognizeOptions.ContentType ?? DetectContentType(receiptFileStream, nameof(receiptFileStream));
 
             ResponseWithHeaders<ServiceAnalyzeReceiptAsyncHeaders> response = ServiceClient.RestClient.AnalyzeReceiptAsync(contentType, receiptFileStream, includeTextDetails: recognizeOptions.IncludeTextContent, cancellationToken);
             return new RecognizeReceiptsOperation(ServiceClient, response.Headers.OperationLocation);
@@ -238,11 +213,9 @@ public virtual RecognizeReceiptsOperation StartRecognizeReceiptsFromUri(Uri rece
         public virtual RecognizeCustomFormsOperation StartRecognizeCustomForms(string modelId, Stream formFileStream, RecognizeOptions recognizeOptions = default, CancellationToken cancellationToken = default)
         {
             recognizeOptions ??= new RecognizeOptions();
+            ContentType contentType = recognizeOptions.ContentType ?? DetectContentType(formFileStream, nameof(formFileStream));
 
-            // TODO: automate content-type detection
-            // https://github.com/Azure/azure-sdk-for-net/issues/10329
-
-            ResponseWithHeaders<ServiceAnalyzeWithCustomModelHeaders> response = ServiceClient.RestClient.AnalyzeWithCustomModel(new Guid(modelId), ContentType.Pdf, formFileStream, includeTextDetails: recognizeOptions.IncludeTextContent, cancellationToken);
+            ResponseWithHeaders<ServiceAnalyzeWithCustomModelHeaders> response = ServiceClient.RestClient.AnalyzeWithCustomModel(new Guid(modelId), contentType, formFileStream, includeTextDetails: recognizeOptions.IncludeTextContent, cancellationToken);
             return new RecognizeCustomFormsOperation(ServiceClient, modelId, response.Headers.OperationLocation);
         }
 
@@ -278,11 +251,9 @@ public virtual RecognizeCustomFormsOperation StartRecognizeCustomFormsFromUri(st
         public virtual async Task<RecognizeCustomFormsOperation> StartRecognizeCustomFormsAsync(string modelId, Stream formFileStream, RecognizeOptions recognizeOptions = default, CancellationToken cancellationToken = default)
         {
             recognizeOptions ??= new RecognizeOptions();
+            ContentType contentType = recognizeOptions.ContentType ?? DetectContentType(formFileStream, nameof(formFileStream));
 
-            // TODO: automate content-type detection
-            // https://github.com/Azure/azure-sdk-for-net/issues/10329
-
-            ResponseWithHeaders<ServiceAnalyzeWithCustomModelHeaders> response = await ServiceClient.RestClient.AnalyzeWithCustomModelAsync(new Guid(modelId), ContentType.Pdf, formFileStream, includeTextDetails: recognizeOptions.IncludeTextContent, cancellationToken).ConfigureAwait(false);
+            ResponseWithHeaders<ServiceAnalyzeWithCustomModelHeaders> response = await ServiceClient.RestClient.AnalyzeWithCustomModelAsync(new Guid(modelId), contentType, formFileStream, includeTextDetails: recognizeOptions.IncludeTextContent, cancellationToken).ConfigureAwait(false);
             return new RecognizeCustomFormsOperation(ServiceClient, modelId, response.Headers.OperationLocation);
         }
 
@@ -318,5 +289,35 @@ public virtual FormTrainingClient GetFormTrainingClient()
         }
 
         #endregion Training client
+
+        /// <summary>
+        /// Used as part of argument validation. Detects the <see cref="ContentType"/> of a stream and
+        /// throws an <see cref="ArgumentException"/> in case of failure.
+        /// </summary>
+        /// <param name="stream">The stream to which the content type detection attempt will be performed.</param>
+        /// <param name="paramName">The original parameter name of the <paramref name="stream"/>. Used to create exceptions in case of failure.</param>
+        /// <returns>The detected <see cref="ContentType"/>.</returns>
+        /// <exception cref="ArgumentException">Happens when detection fails or cannot be performed.</exception>
+        private static ContentType DetectContentType(Stream stream, string paramName)
+        {
+            ContentType contentType;
+
+            if (!stream.CanSeek)
+            {
+                throw new ArgumentException("Content type cannot be detected because stream is not seekable.", paramName);
+            }
+
+            if (!stream.CanRead)
+            {
+                throw new ArgumentException("Content type cannot be detected because stream is not readable.", paramName);
+            }
+
+            if (!stream.TryGetContentType(out contentType))
+            {
+                throw new ArgumentException("Content type of the stream could not be detected.", paramName);
+            }
+
+            return contentType;
+        }
     }
 }

sdk/formrecognizer/Azure.AI.FormRecognizer/src/FormRecognizerClientOptions.cs

@@ -35,7 +35,6 @@ public enum ServiceVersion
 #pragma warning restore CA1707 // Identifiers should not contain underscores
         }
 
-
         /// <summary>
         /// </summary>
         public ServiceVersion Version { get; }

sdk/formrecognizer/Azure.AI.FormRecognizer/src/ReceiptExtensions.cs

@@ -1,10 +1,6 @@
 // Copyright (c) Microsoft Corporation. All rights reserved.
 // Licensed under the MIT License.
 
-using System;
-using System.Collections.Generic;
-using System.Text;
-
 namespace Azure.AI.FormRecognizer.Models
 {
     /// <summary>

sdk/formrecognizer/Azure.AI.FormRecognizer/src/RecognizeOptions.cs

@@ -17,5 +17,8 @@ public RecognizeOptions()
         /// </summary>
         public bool IncludeTextContent { get; set; } = false;
 
+        /// <summary>
+        /// </summary>
+        public ContentType? ContentType { get; set; } = null;
     }
 }

sdk/formrecognizer/Azure.AI.FormRecognizer/src/StreamExtensions.cs

@@ -0,0 +1,102 @@
+// Copyright (c) Microsoft Corporation. All rights reserved.
+// Licensed under the MIT License.
+
+using System;
+using System.IO;
+using Azure.AI.FormRecognizer.Models;
+
+namespace Azure.AI.FormRecognizer
+{
+    internal static class StreamExtensions
+    {
+        /// <summary>The set of bytes expected to be present at the start of PDF files.</summary>
+        private static byte[] PdfHeader = new byte[] { 0x25, 0x50, 0x44, 0x46 };
+
+        /// <summary>The set of bytes expected to be present at the start of PNG files.</summary>
+        private static byte[] PngHeader = new byte[] { 0x89, 0x50, 0x4E, 0x47 };
+
+        /// <summary>The set of bytes expected to be present at the start of JPEG files.</summary>
+        private static byte[] JpegHeader = new byte[] { 0xFF, 0xD8 };
+
+        /// <summary>The set of bytes expected to be present at the start of TIFF (little-endian) files.</summary>
+        private static byte[] TiffLeHeader = new byte[] { 0x49, 0x49, 0x2A, 0x00 };
+
+        /// <summary>The set of bytes expected to be present at the start of TIFF (big-endian) files.</summary>
+        private static byte[] TiffBeHeader = new byte[] { 0x4D, 0x4D, 0x00, 0x2A };
+
+        /// <summary>
+        /// Attemps to detect the <see cref="ContentType"/> of a stream of bytes. The algorithm searches through
+        /// the first set of bytes in the stream and compares it to well-known file signatures.
+        /// </summary>
+        /// <param name="stream">The stream to which the content type detection attempt will be performed.</param>
+        /// <param name="contentType">If the detection is successful, outputs the detected content type. Otherwise, <c>default</c>.</param>
+        /// <returns><c>true</c> if the detection was successful. Otherwise, <c>false</c>.</returns>
+        /// <exception cref="NotSupportedException">Happens when <paramref name="stream"/> is not seekable or readable.</exception>
+        public static bool TryGetContentType(this Stream stream, out ContentType contentType)
+        {
+            var originalPosition = stream.Position;
+            var bytesCount = Math.Min(stream.Length, PdfHeader.Length);
+
+            bool hasFoundType = false;
+            contentType = default;
+
+            bool couldBePdf = true;
+            bool couldBePng = true;
+            bool couldBeJpeg = true;
+            bool couldBeTiffLe = true;
+            bool couldBeTiffBe = true;
+
+            Func<byte[], int, bool> isAtEnd = (array, index) => index == array.Length - 1;
+            Func<byte[], int, byte, bool> hasByteAt = (array, index, expected) => index < array.Length && array[index] == expected;
+
+            for (int i = 0; i < bytesCount; i++)
+            {
+                byte currentByte = (byte)stream.ReadByte();
+
+                couldBePdf = couldBePdf && hasByteAt(PdfHeader, i, currentByte);
+                if (couldBePdf && isAtEnd(PdfHeader, i))
+                {
+                    contentType = ContentType.Pdf;
+                    hasFoundType = true;
+                    break;
+                }
+
+                couldBePng = couldBePng && hasByteAt(PngHeader, i, currentByte);
+                if (couldBePng && isAtEnd(PngHeader, i))
+                {
+                    contentType = ContentType.Png;
+                    hasFoundType = true;
+                    break;
+                }
+
+                couldBeJpeg = couldBeJpeg && hasByteAt(JpegHeader, i, currentByte);
+                if (couldBeJpeg && isAtEnd(JpegHeader, i))
+                {
+                    contentType = ContentType.Jpeg;
+                    hasFoundType = true;
+                    break;
+                }
+
+                couldBeTiffLe = couldBeTiffLe && hasByteAt(TiffLeHeader, i, currentByte);
+                if (couldBeTiffLe && isAtEnd(TiffLeHeader, i))
+                {
+                    contentType = ContentType.Tiff;
+                    hasFoundType = true;
+                    break;
+                }
+
+                couldBeTiffBe = couldBeTiffBe && hasByteAt(TiffBeHeader, i, currentByte);
+                if (couldBeTiffBe && isAtEnd(TiffBeHeader, i))
+                {
+                    contentType = ContentType.Tiff;
+                    hasFoundType = true;
+                    break;
+                }
+            }
+
+            stream.Position = originalPosition;
+
+            return hasFoundType;
+        }
+    }
+}