From 5d14ac74f614631a38be95cc9724ff38068c6c76 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Justin=20Nu=C3=9F?= <nuss.justin@gmail.com>
Date: Wed, 2 Aug 2017 19:18:30 +0200
Subject: [PATCH] encoding/csv: report line start line in errors

Errors returned by Reader contain the line where the Reader originally
encountered the error. This can be suboptimal since that line does not
always correspond with the line the current record/field started at.

This can easily happen with LazyQuotes as seen in #19019, but also
happens for example when a quoted fields has no closing quote and
the parser hits EOF before it finds another quote.

When this happens finding the erroneous field can be somewhat
complicated and time consuming, and in most cases it would be better to
report the line where the record started.

This change updates Reader to keep track of the line on which a record
begins and uses it for errors instead of the current line, making it
easier to find errors.

Although a user-visible change, this should have no impact on existing
code, since most users don't explicitly work with the line in the error
and probably already expect the new behaviour.

Updates #19019

Change-Id: Ic9bc70fad2651c69435d614d537e7a9266819b05
Reviewed-on: https://go-review.googlesource.com/52830
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
---
 src/encoding/csv/reader.go      | 11 +++++++----
 src/encoding/csv/reader_test.go | 14 ++++++++++++++
 2 files changed, 21 insertions(+), 4 deletions(-)

diff --git a/src/encoding/csv/reader.go b/src/encoding/csv/reader.go
index a3497c84f9..4085321a71 100644
--- a/src/encoding/csv/reader.go
+++ b/src/encoding/csv/reader.go
@@ -115,9 +115,10 @@ type Reader struct {
 	// By default, each call to Read returns newly allocated memory owned by the caller.
 	ReuseRecord bool
 
-	line   int
-	column int
-	r      *bufio.Reader
+	line       int
+	recordLine int // line where the current record started
+	column     int
+	r          *bufio.Reader
 	// lineBuffer holds the unescaped fields read by readField, one after another.
 	// The fields can be accessed by using the indexes in fieldIndexes.
 	// Example: for the row `a,"b","c""d",e` lineBuffer will contain `abc"de` and
@@ -142,7 +143,7 @@ func NewReader(r io.Reader) *Reader {
 // error creates a new ParseError based on err.
 func (r *Reader) error(err error) error {
 	return &ParseError{
-		Line:   r.line,
+		Line:   r.recordLine,
 		Column: r.column,
 		Err:    err,
 	}
@@ -251,7 +252,9 @@ func (r *Reader) parseRecord(dst []string) (fields []string, err error) {
 	// Each record starts on a new line. We increment our line
 	// number (lines start at 1, not 0) and set column to -1
 	// so as we increment in readRune it points to the character we read.
+	// We track the line where the record starts in recordLine for use in errors.
 	r.line++
+	r.recordLine = r.line
 	r.column = -1
 
 	// Peek at the first rune. If it is an error we are done.
diff --git a/src/encoding/csv/reader_test.go b/src/encoding/csv/reader_test.go
index 5ab1b61256..76e94bab3e 100644
--- a/src/encoding/csv/reader_test.go
+++ b/src/encoding/csv/reader_test.go
@@ -270,6 +270,20 @@ x,,,
 			{"c", "d"},
 		},
 	},
+	{ // issue 19019
+		Name:   "RecordLine1",
+		Input:  "a,\"b\nc\"d,e",
+		Error:  `extraneous " in field`,
+		Line:   1,
+		Column: 1,
+	},
+	{
+		Name:   "RecordLine2",
+		Input:  "a,b\n\"d\n\n,e",
+		Error:  `extraneous " in field`,
+		Line:   2,
+		Column: 2,
+	},
 }
 
 func TestRead(t *testing.T) {