Indexing DataFrame columns can return empty DataFrame

[juliadata]

Indexing DataFrame columns can return empty DataFrame

[tshort]

That's the same error that Doug brought up a couple of days ago, right? Harlan thinks it's because of a change in Julia indexing. I don't get that on a slightly older version of Julia.

[HarlanH]

Thanks for posting this, Chris. Yes, this is the error due to a change in the way the : operator works that happened a few days ago. Haven't heard back from Jeff -- I'll email him again...

[HarlanH]

Jeff sayeth: df[1,:] will now get converted to df[1,1:div(length(df),size(df,1))].

[tshort]

Do we have to redefine length() for a DF? Can we override the colon() function?

[HarlanH]

Oh, bleah. I hadn't thought this through when I copy-pasted it. No, we can't over-ride this, it's part of the parser. I'm going to file a bug report.

[HarlanH]

OK, bug report here: JuliaLang/julia#1066 If you've got a minute, would you mind writing and checking in the pointless length(AbstractDataFrame) function? That'll work around this problem.

[tshort]

I think I got that in.

But, in the process, I ended up pushing/closing this pull request:

https://github.com/HarlanH/JuliaData/pull/37

If someone knows how to undo that, please do. That could use review and comment.

And, I thought I was just starting to figure out git and github... Grrr.

[tshort]

More problems. I had to undefine numel for DataFrames. It turns out that numel is a direct alias for length, so if I redefine numel, I redefine length. I don't think I like that feature.

[tshort]

Another problem: d[:] doesn't work. That should probably return the whole DF.