Skip the whole record after a field size error

lib/csv/parser.rb

@@ -942,7 +942,11 @@ def skip_line?(line)
     def validate_field_size(field)
       return unless @max_field_size
       return if field.size <= @max_field_size
-      ignore_broken_line
+      if block_given?
+        yield
+      else
+        ignore_broken_line
+      end
       message = "Field size exceeded: #{field.size} > #{@max_field_size}"
       raise MalformedCSVError.new(message, @lineno)
     end
@@ -1055,7 +1059,7 @@ def parse_quotable_robust(&block)
         value = parse_column_value
         if value
           @scanner.scan_all(@strip_value) if @strip_value
-          validate_field_size(value)
+          validate_field_size(value) {ignore_broken_record}
         end
         if parse_column_end
           row << value
@@ -1264,6 +1268,20 @@ def ignore_broken_line
       @lineno += 1
     end
 
+    def ignore_broken_record
+      lineno = @lineno
+      begin
+        while parse_column_end
+          @scanner.scan_all(@strip_value) if @strip_value
+          parse_column_value
+          @scanner.scan_all(@strip_value) if @strip_value
+        end
+        ignore_broken_line unless parse_row_end
+      rescue MalformedCSVError
+      end
+      @lineno = lineno + 1
+    end
+
     def start_row
       if @last_line
         @last_line = nil

test/csv/parse/test_general.rb

@@ -329,6 +329,83 @@ def test_max_field_size_in_extended_column_exceeding
     assert_parse_errors_out(data, max_field_size: 4)
   end
 
+  def test_max_field_size_multiline_record_liberal_parsing
+    csv = CSV.new(%Q{abcde,"x\ny"\nc,d\n},
+                  max_field_size: 4, liberal_parsing: true)
+    error = assert_raise(CSV::MalformedCSVError) do
+      csv.shift
+    end
+    assert_equal("Field size exceeded: 5 > 4 in line 1.",
+                 error.message)
+    assert_equal(%Q{abcde,"x\ny"\n}, csv.line)
+    assert_equal(["c", "d"], csv.shift)
+    assert_equal(2, csv.lineno)
+  end
+
+  def test_max_field_size_three_line_record_liberal_parsing
+    csv = CSV.new(%Q{abcde,"x\ny\nz"\nc,d\n},
+                  max_field_size: 4, liberal_parsing: true)
+    error = assert_raise(CSV::MalformedCSVError) do
+      csv.shift
+    end
+    assert_equal("Field size exceeded: 5 > 4 in line 1.",
+                 error.message)
+    assert_equal(%Q{abcde,"x\ny\nz"\n}, csv.line)
+    assert_equal(["c", "d"], csv.shift)
+    assert_equal(2, csv.lineno)
+  end
+
+  def test_max_field_size_multiline_record_after_multiline_row
+    csv = CSV.new(%Q{"a\nb",c\nabcde,"x\ny"\nc,d\n}, max_field_size: 4)
+    assert_equal(["a\nb", "c"], csv.shift)
+    error = assert_raise(CSV::MalformedCSVError) do
+      csv.shift
+    end
+    assert_equal("Field size exceeded: 5 > 4 in line 2.",
+                 error.message)
+    assert_equal(%Q{abcde,"x\ny"\n}, csv.line)
+    assert_equal(["c", "d"], csv.shift)
+    assert_equal(3, csv.lineno)
+  end
+
+  def test_max_field_size_multiline_record_strip
+    csv = CSV.new(%Q{abcde , "x\ny" \nc,d\n},
+                  max_field_size: 4, strip: true)
+    error = assert_raise(CSV::MalformedCSVError) do
+      csv.shift
+    end
+    assert_equal("Field size exceeded: 5 > 4 in line 1.",
+                 error.message)
+    assert_equal(%Q{abcde , "x\ny" \n}, csv.line)
+    assert_equal(["c", "d"], csv.shift)
+    assert_equal(2, csv.lineno)
+  end
+
+  def test_max_field_size_value_after_quoted_field
+    csv = CSV.new(%Q{"abcde"junk,b\nc,d\n}, max_field_size: 4)
+    error = assert_raise(CSV::MalformedCSVError) do
+      csv.shift
+    end
+    assert_equal("Field size exceeded: 5 > 4 in line 1.",
+                 error.message)
+    assert_equal(%Q{"abcde"junk,b\n}, csv.line)
+    assert_equal(["c", "d"], csv.shift)
+    assert_equal(2, csv.lineno)
+  end
+
+  def test_max_field_size_unclosed_quote_liberal_parsing
+    csv = CSV.new(%Q{abcde,"x\nc,d\n},
+                  max_field_size: 4, liberal_parsing: true)
+    error = assert_raise(CSV::MalformedCSVError) do
+      csv.shift
+    end
+    assert_equal("Field size exceeded: 5 > 4 in line 1.",
+                 error.message)
+    assert_equal(%Q{abcde,"x\nc,d\n}, csv.line)
+    assert_nil(csv.shift)
+    assert_equal(1, csv.lineno)
+  end
+
   def test_row_sep_auto_cr
     assert_equal([["a"]], CSV.parse("a\r"))
   end