From f26b3d2205247cefb1d132372f15ce9c48acb504 Mon Sep 17 00:00:00 2001
From: FARHAD SHAKERIN <fxs130430@utdallas.edu>
Date: Sun, 4 Dec 2022 22:18:12 -0800
Subject: [PATCH] [FR DateTimeV2] Stack overflow fix in .NET (#3050)

---
 .../Parsers/BaseDurationParser.cs             | 11 +++++++-
 Specs/DateTime/French/DateTimeModel.json      | 27 ++++++++++++++++---
 2 files changed, 34 insertions(+), 4 deletions(-)

diff --git a/.NET/Microsoft.Recognizers.Text.DateTime/Parsers/BaseDurationParser.cs b/.NET/Microsoft.Recognizers.Text.DateTime/Parsers/BaseDurationParser.cs
index 9f08ef6d2e..3088611999 100644
--- a/.NET/Microsoft.Recognizers.Text.DateTime/Parsers/BaseDurationParser.cs
+++ b/.NET/Microsoft.Recognizers.Text.DateTime/Parsers/BaseDurationParser.cs
@@ -398,7 +398,16 @@ private DateTimeResolutionResult ParseMergedDuration(string text, DateObject ref
             var durationExtractor = this.config.DurationExtractor;
 
             // DurationExtractor without parameter will not extract merged duration
-            var ers = durationExtractor.Extract(text, referenceTime);
+
+            // TrimStart() was added to address a bug with french duration expression "depuis ans"
+            // for which the basecase of the recursive call (i.e., if(ers.Count <= 1))
+            // would never be reached in which case the stack would overflow.
+            // The statement if(minStart){...} is meant to find the isolated unit as explained in
+            // the below comment. However, if there is whitespace before the extacted entity
+            // (as in " ans") the minStart will be greater than 1 and the Followed Unit regex
+            // keeps on matching with "ans" and it adds it to ers and it always has
+            // more than one item in it, hence the recursion never ends.
+            var ers = durationExtractor.Extract(text.TrimStart(), referenceTime);
 
             // If the duration extractions do not start at 0, check if the input starts with an isolated unit.
             // This happens for example with patterns like "next week and 3 days" where "next" is not part of the extraction.
diff --git a/Specs/DateTime/French/DateTimeModel.json b/Specs/DateTime/French/DateTimeModel.json
index 992cbe1aa7..4e608dcb21 100644
--- a/Specs/DateTime/French/DateTimeModel.json
+++ b/Specs/DateTime/French/DateTimeModel.json
@@ -3624,7 +3624,7 @@
         "End": 67
       }
     ]
-  },  
+  },
   {
     "Input": "Je vais rentrer le 4 janvier 2019.",
     "NotSupported": "dotnet, javascript, python, java",
@@ -8694,6 +8694,27 @@
       }
     ]
   },
+  {
+    "Input": "bonjour la maison n'est plus habitée depuis ans",
+    "Results": [
+      {
+        "Text": "depuis ans",
+        "Start": 37,
+        "End": 46,
+        "TypeName": "datetimeV2.duration",
+        "Resolution": {
+          "values": [
+            {
+              "timex": "P1Y",
+              "Mod": "since",
+              "type": "duration",
+              "value": "31536000"
+            }
+          ]
+        }
+      }
+    ]
+  },
   {
     "Input": "Rechercher les enregistrements durent moins de 2 heures ou plus de 4 jours et pas moins de 30 minutes.",
     "NotSupported": "dotnet, javascript, python, java",
@@ -21372,7 +21393,7 @@
               "type": "time",
               "value": "08:00:00"
             },
-			{
+            {
               "timex": "T20",
               "type": "time",
               "value": "20:00:00"
@@ -21400,7 +21421,7 @@
               "type": "time",
               "value": "01:00:00"
             },
-			{
+            {
               "timex": "T13",
               "type": "time",
               "value": "13:00:00"