fix(commons): OverrideCsvParser — preserve literal mid-field quotes, error on unterminated quoted field (T16 CSV)

src/ZB.MOM.WW.ScadaBridge.Commons/Types/OverrideCsvParser.cs

@@ -58,7 +58,11 @@ public static class OverrideCsvParser
             if (string.IsNullOrWhiteSpace(rawLine))
                 continue;
 
-            var fields = SplitFields(rawLine);
+            if (!SplitFields(rawLine, out var fields))
+            {
+                errors.Add($"Line {lineNumber}: Unterminated quoted field.");
+                continue;
+            }
 
             if (!headerSeen)
             {
@@ -130,16 +134,30 @@ public static class OverrideCsvParser
     private static string? NullIfEmpty(string field) => field.Length == 0 ? null : field;
 
     /// <summary>
-    /// RFC-4180-ish field splitter for a single physical line: quoted fields may
+    /// RFC-4180-ish field splitter for a single physical line. Quoting rules:
-    /// embed commas and doubled quotes (<c>""</c> → <c>"</c>); unquoted fields are
+    /// <list type="bullet">
-    /// whitespace-trimmed.
+    /// <item>A field is <i>quoted</i> only if its first non-whitespace char is a
+    /// <c>"</c> (leading whitespace before the opening quote is allowed and
+    /// ignored). Inside a quoted field, commas are literal and <c>""</c> is an
+    /// escaped single quote; the closing <c>"</c> must be the last non-whitespace
+    /// char of the field (trailing whitespace after the close is allowed and
+    /// ignored).</item>
+    /// <item>A <c>"</c> appearing anywhere else in an unquoted field (i.e. after
+    /// non-whitespace content) is a <b>literal</b> character and is preserved.</item>
+    /// <item>Unquoted fields are whitespace-trimmed; quoted field values are kept
+    /// verbatim.</item>
+    /// </list>
+    /// Returns <c>true</c> with the split <paramref name="fields"/> on success;
+    /// returns <c>false</c> when a quoted field is opened but never closed before
+    /// end-of-line (the caller emits a per-line "unterminated" error).
     /// </summary>
-    private static List<string> SplitFields(string line)
+    private static bool SplitFields(string line, out List<string> fields)
     {
-        var fields = new List<string>();
+        fields = new List<string>();
         var field = new System.Text.StringBuilder();
-        var inQuotes = false;
+        var inQuotes = false;   // currently between an opening and closing quote
-        var quoted = false; // this field was (at least partly) quoted → preserve whitespace
+        var quoted = false;     // this field opened with a quote → keep value verbatim
+        var sawContent = false; // any non-whitespace char seen in the current field yet
 
         for (var i = 0; i < line.Length; i++)
         {
@@ -157,7 +175,7 @@ public static class OverrideCsvParser
                     }
                     else
                     {
-                        inQuotes = false;
+                        inQuotes = false; // closing quote; only trailing whitespace may follow
                     }
                 }
                 else
@@ -170,23 +188,40 @@ public static class OverrideCsvParser
 
             switch (c)
             {
-                case '"':
-                    inQuotes = true;
-                    quoted = true;
-                    break;
                 case ',':
                     fields.Add(Finalize(field, quoted));
                     field.Clear();
+                    inQuotes = false;
                     quoted = false;
+                    sawContent = false;
+                    break;
+                case '"' when !sawContent:
+                    // Opening quote: first non-whitespace char of the field. Any
+                    // leading whitespace seen so far is part of the (ignored) prefix.
+                    field.Clear();
+                    inQuotes = true;
+                    quoted = true;
+                    sawContent = true;
                     break;
                 default:
+                    // After a quoted field has closed, only whitespace may appear
+                    // before the next delimiter — it is ignored, not appended.
+                    if (quoted)
+                        break;
+
+                    // A '"' here (sawContent already true) falls through as a literal.
+                    if (!char.IsWhiteSpace(c))
+                        sawContent = true;
                     field.Append(c);
                     break;
             }
         }
 
+        if (inQuotes)
+            return false; // opened a quoted field that was never closed
+
         fields.Add(Finalize(field, quoted));
-        return fields;
+        return true;
     }
 
     private static string Finalize(System.Text.StringBuilder field, bool quoted)

tests/ZB.MOM.WW.ScadaBridge.Commons.Tests/OverrideCsvParserTests.cs

@@ -169,4 +169,80 @@ public class OverrideCsvParserTests
         Assert.Equal("42", result.Rows[0].Value);
         Assert.Equal("  spaced  ", result.Rows[1].Value);
     }
+
+    [Fact]
+    public void Parse_MidFieldUnquotedQuote_IsPreservedAsLiteral()
+    {
+        // A '"' that does NOT open a field (it follows non-whitespace content in an
+        // unquoted field) is a literal character — 'va"lue' must survive intact.
+        const string csv = "AttributeName,Value,ElementType\nName,va\"lue,Type\n";
+
+        var result = OverrideCsvParser.Parse(csv);
+
+        Assert.Empty(result.Errors);
+        var row = Assert.Single(result.Rows);
+        Assert.Equal("Name", row.AttributeName);
+        Assert.Equal("va\"lue", row.Value);
+        Assert.Equal("Type", row.ElementType);
+    }
+
+    [Fact]
+    public void Parse_UnterminatedQuotedField_ProducesLineNumberedErrorAndExcludesRow()
+    {
+        // A quote opens the field but is never closed before end-of-line → malformed.
+        const string csv = "AttributeName,Value\nName,\"unclosed\n";
+
+        var result = OverrideCsvParser.Parse(csv);
+
+        Assert.Empty(result.Rows);
+        var error = Assert.Single(result.Errors);
+        Assert.Contains("2", error);
+        Assert.Contains("Unterminated", error, StringComparison.OrdinalIgnoreCase);
+    }
+
+    [Fact]
+    public void Parse_WellFormedQuotedFieldWithComma_StillParses()
+    {
+        // Regression guard alongside the unterminated-quote fix: a properly closed
+        // quoted field embedding a comma must still round-trip.
+        const string csv = "AttributeName,Value,ElementType\nName,\"a,b\",Type\n";
+
+        var result = OverrideCsvParser.Parse(csv);
+
+        Assert.Empty(result.Errors);
+        var row = Assert.Single(result.Rows);
+        Assert.Equal("Name", row.AttributeName);
+        Assert.Equal("a,b", row.Value);
+        Assert.Equal("Type", row.ElementType);
+    }
+
+    [Fact]
+    public void Parse_EmptyQuotedField_YieldsNullValue()
+    {
+        // A bare "" is an empty quoted field; empty → null per the empty-field rule.
+        const string csv = "AttributeName,Value,ElementType\nName,\"\",Type\n";
+
+        var result = OverrideCsvParser.Parse(csv);
+
+        Assert.Empty(result.Errors);
+        var row = Assert.Single(result.Rows);
+        Assert.Equal("Name", row.AttributeName);
+        Assert.Null(row.Value);
+        Assert.Equal("Type", row.ElementType);
+    }
+
+    [Fact]
+    public void Parse_QuotedFieldWithTrailingWhitespaceAfterClose_IsAccepted()
+    {
+        // The closing quote may be followed by ignorable trailing whitespace before
+        // the delimiter; the field value itself is preserved verbatim.
+        const string csv = "AttributeName,Value,ElementType\nName,\"a,b\"  ,Type\n";
+
+        var result = OverrideCsvParser.Parse(csv);
+
+        Assert.Empty(result.Errors);
+        var row = Assert.Single(result.Rows);
+        Assert.Equal("a,b", row.Value);
+        Assert.Equal("Type", row.ElementType);
+    }
 }