jeffpaine / csvsplit Goto Github PK

Seems a little cluttered

To be more idiomatically go, the package documentation ought to be in comments directly above the package main declaration. The only concern I have is with either having an anemic README.md or having full versions of both but having to keep them in sync.

It should count the lines in a file and pad filenames appropriately. e.g. 001.csv, 002.csv instead of 1.csv, 2.csv. This should default to false as it will require a file to be read twice.

e.g instead of -i we should use -input.

This part especially feels wonky

if _, err := os.Stat(fileName); os.IsNotExist(err)...

Add one to each file?

Right now, the header (if it's present in the original) is only printed to the first file.

Or at least make it optional? E.g.

$ csvsplit -records 10 file.csv

Instead of / equivalent to

$ csvsplit -records 10 -input files.csv

Right now it's 1000, should we prompt for it, instead?

Right now, we start with a zero length slice and keep appending to it. https://github.com/JeffPaine/csvsplit/blob/master/main.go#L93

I wonder if it would perform better as a pre-sized array?

Depends on #18, so we can measure our performance before and after.

Should be able to use len() on slice.

Seems more logical

Right now, we auto-append 001, 002, etc. to output filenames. It would be cool to have this zero padding reflect the number of files actually saved (e.g. 01.csv for < 99 files), but I would think that would require us to count (read into memory) all the lines of the file, first.

e.g. -r and --records ?