I don't understand how this would be used. Can you show an example input file, and example code to read that file? Or just write the docs for the IndexFields method? It seems to me that this can only read different orderings of "foo", "bar", and "baz", which seems useless. Probably I misunderstand.

Sorry, I didn't see this before. Here is a playground link with a more complete usage example and a first pass at function documentation.

The usecase for this is that you have a CSV file, which you know has certain headers, but you either don't know which index they are at, or don't want to hardcode those indices, in case they change in the future. Obviously, you could just write this yourself because it's a short function that doesn't use csv.Reader internals, but I think it smooths out a common occurrence.

I often see CSV parsing code with magic numbers like record[3] + " <" + record[5] + ">". You can improve that code significantly by writing record[nameIdx] + " <" + record[emailIdx] + ">". To do that you can hardcode const nameIdx, emailIdx = 3, 5, but if the file already has a header, why not use that to get the indices?

It occurs to me that it would be a nicer interface to take in a struct and use reflection to set the values. I don't have time ATM to whip up a demo, but usage like:

    in := `first_name,last_name,username
"Rob","Pike",rob
Ken,Thompson,ken
"Robert","Griesemer","gri"
`
    r := csv.NewReader(strings.NewReader(in))
    var idxs struct { 
        Username int 
    }
    if err := r.IndexFields(&idxs); err != nil {
        log.Fatal(err)
    }

    // Print all usernames
    for {
        record, err := r.Read()
        if err == io.EOF {
            break
        }
        if err != nil {
            log.Fatal(err)
        }

        fmt.Println(record[idxs.Username])
    }

Thanks for the example. There are a hundred ways we could extend the encoding/csv package, to account for all the many different ways that people use CSV. Our preference has been to stick to RFC 4180. It's easy enough to implement your suggestion outside of the encoding/csv package, so my vote would be to do that rather than to extend the standard package. See https://golang.org/doc/faq#x_in_std .

There is a somewhat nicer solution to the problem (error handling and implementation left as exercise):

data := mycsv.NewReader(input)
firstname := data.String("first_name")
lastname := data.String("last_name")
username := data.String("username")
age := data.Int("age")

for data.Next() {
	fmt.Println(*firstname, *lastname, *username, *age)
}

But, I agree with Ian, that this probably doesn't belong in standard package.

Per discussion above, let's leave this for add-on packages that can more easily customize to very specific use cases.