CSVKit: CSV utilities that includes csvsql, csvgrep, csvstat, and more

> You will already have this program in your Linux distro

Yum disagrees on CentOS 6.4.

I tried out OpenRefine previously on the large CSV's I have to deal with at work largely on the recommendation of HN comments. On small stuff to medium size stuff, it's pretty nice. But for larger sets (+1GB) it starts to slow down and eventually will fail to commit changes when the set is big enough.

The only answer I've found to consistently work quickly, with the ability to explore the data, is the killer combo of iPython Notebook + pandas' read_csv[0] + a lot of RAM -- 10GB CSV on disk becomes ~20GB in memory (don't know why yet). When I say quick, I mean 10GB CSV un-cached disk to RAM in <5Min including fuzzy parsing on dates.

The nice part is, when you have things figured out, you can enable a chunked reading to get back in-core on machines of lesser specs. Further, you can dump the pandas DataFrame to HDF, thereafter having ludicrous-speed IO & 'where' queries.

Still though, OpenRefine is much more turn key and feature rich.

[0]http://pandas.pydata.org/pandas-docs/version/0.13.1/generate...

Why? I know and love csvkit, but never heard of csvfix.

That depends on what you want to do with your junky input. And if it's junky enough, no tool is going to be able to divine what you want to do with the data until you do some manual cleanup. However, just to put it out there, csvkit has a utility, csvclean, that will split a file based on which lines are valid csv and which are not, allowing you to more quickly dive into what's wrong with the bad lines.

They're not replacements for those tools. They're new tools that work with a specific tabular data format.

Grep understands text files, not CSV syntax. They're not equivalent. For example, the following CSV document has two lines, but grep treats them as three:

  1,2,5,"Hello"
  3,4,5,"How are
  you doing?"

And don't forget AWK!

I've been playing with these old school tools lately on medium sized tabular data files and I've been impressed by the performance and capabilities.

> The syntax doesn't seem noticeably clearer?

From the very first example:

    csvgrep -c 1 -m ILLINOIS 

using coreutils grep you'd need something like this:

    grep -E '^ILLINOIS,' 2009.csv 

Oh, wait, that doesn't work because the file was last saved by something which quotes it.

    grep -E '^"?ILLINOIS"?,'

except that would produce unexpected results with 'ILLINOIS"' so it really needs to be something like:

    grep -E '^(["]{0,1})ILLINOIS\1,'

Bear in mind that this is the simplest possible case and doesn't even touch on issues like quoting in the shell or needing to handle files which have embedded separators as data values (imagine what happens when our erstwhile grep data-miner needs to check the stats for `WASHINGTON, DISTRICT OF COLUMBIA`…).

In all but the most trivial cases it's safer and easier to use tools designed for the job. csvkit also has the very nice property that it's callable as a Python library so when you outgrow a simple shell processing pipeline you could migrate your notes / scripts to a full Python program without having to retest everything related to basic file I/O, unicode, etc. which you would otherwise need to do when switching readers.

(Bear in mind that the author works in data journalism – the target user is not a grizzled Unix expert but someone who has a bunch of CSV files and a full time job which is not shell scripting)

> And - without testing - I presume csvkit in Python is a bit slower than the GNU coreutils in C?

Perhaps but it'd be unlikely for it to be noticeable for n less than millions on a remotely modern system – the Python regular expression engine has quite decent performance and if that became an issue, PyPy will even JIT them for you. In the very few cases I've seen where there is a noticeable difference it was always because the Python version was decoding Unicode and the shell tools were running with LC_ALL=C, which meant that corrupt data made it further before being caught and, in some cases, either failing to match all of the records or subtly quietly things by not extracting all of a combined character, etc.

For the target use-case, however, this is likely all to be many, many orders of magnitude less than the time most people would spend debugging regex patterns.