encoding/csv: getting a line number of current record when using csv.Reader

[alser777]

Please answer these questions before submitting your issue. Thanks!

What version of Go are you using (go version)?

1.10.3

Does this issue reproduce with the latest release?

1.10.3 is the latest

What did you do?

I have used csv.Reader to parse a CSV file and I wanted to report an error to user in case some record contained invalid data. I also wanted to include a line number of invalid record in an error message.

What did you expect to see?

some method that would return a line number where current record starts

What did you see instead?

There is no way to get a line number of current record.

[agnivade]

Is there any issue using a counter variable to keep track of the no. of lines parsed ?

[nussjustin]

@agnivade Probably, since a single record can span multiple lines. E.g. the 3rd record could be from line 7 to 20 in the source.

The Reader only tracks the current line not the line where a record started. We could add a method that returns the current line number, but to get the line where the last record started we would need to change the Reader to remember the line.

[agnivade]

/cc @dsnet

[dsnet]

What's wrong with the current information provided:

type ParseError struct {
        StartLine int   // Line where the record starts
        Line      int   // Line where the error occurred
        Column    int   // Column (rune index) where the error occurred
        Err       error // The actual error
}

Isn't ParseError.StartLine exactly what is being asked for?

[alser777]

@dsnet No, I ask for a method to get StartLine when there is no parse error

[dsnet]

Ah, I see. This seems to be a relatively niche feature, and I'm not sure if the expansion of API is worth it.

Something along the lines of what @agnivade suggested should work. The issue of multi-line records can be resolved by sampling the current line before each call to csv.Reader.Read. However, this is complicated by the fact that csv.Reader always wraps the input reader with a bufio.Reader regardless of whether the input is already a bufio.Reader or not. Thus, any external accounting is invalidated by the internal buffering of the csv.Reader.

If we changed the internal implementation of csv.Reader to avoid re-wrapping a bufio.Reader, you could account for this yourself by creating a io.Reader that counts the number of lines subtracted by the number of newlines in the bufio.Reader buffer (see Peek).

[agnivade]

@dsnet - It is a bit unclear on what is to be done here. Are you suggesting we should check the reader and avoid the re-wrap ?

[agnivade]

ping @dsnet .