trac user alefnula wrote on 2010-09-15

I had a mistake in the example csv, it should look like:

"This is my text, that has a comma inside","Other value","3"[[BR]]
"Another text, with coma","More text, with comma","5"[[BR]]
...

Attachment added by trac user alefnula on 2010-09-15: patch.diff

@rgommers wrote on 2011-03-31

A unit test would be helpful.

Will this issue be fixed?

@jseidel Want to open an PR?

@mattip is this an issue I can work on?

We don't usually assign issues, the rule is that the first person to file a PR should link back to the issue to indicate they are giving it a try. Note that the patch is a broken link so you will have to recreate it. Tests are of course critical,

Is there a reason to use np.genfromtxt() over pandas' pd.read_csv() in 2019?

@ddasilva I'm not sure if this counts as a reason, but one advantage could be that the output is a ndarray rather than a pd.Dataframe. I do see your point though sinice pd.read_csv() does seem to be the clear choice over np.genfromtxt() for most people

@hypercubestart @eric-wieser is this issue still active ?

@tinaoberoi not sure, but from the discussion here #14577 it feels like the general sentiment is to just use pandas.read_csv

I would like to work on this issue, I followed the earlier PRs and the discussion. I agree with @hypercubestart suggestion of:
adding a reference of pandas.read_csv
I would like to make sure before adding the PR.
@rgommers @mattip

@tinaoberoi yes that seems like a good solution. genfromtxt is quite complex and basically beyond rescue. So just saying something in the Notes section like "for data with commas inside quoted strings or other corner cases that genfromtxt struggles with, using pandas.read_csv is a good alternative".

Longer-term, @WarrenWeckesser is working on a better text reader for numpy.