[bug] allow to specify fields in csv writer

[yanjost]

request : allow to specify the field names when writing to a csv file

currently the code uses "context.get_input_fields()". Either I did not understand how it's supposed to be used, or it's a bug !

I will submit a patch

[hartym]

Thanks for the report.

It seems that although sphinx (documentation website) displays the incorrect version of the code, this is something apparently fixed in develop branch (see https://github.com/python-bonobo/bonobo/blob/develop/bonobo/nodes/io/csv.py#L108). Reading the code (and docs), the fields option should do the trick.

Can you confirm it looks like it is, and that it fixes your problem on your side?

Thanks!

[yanjost]

Thanks for the speedy response ! It seems like the fix in develop branch is very similar to mine, with the skip_header option added. Is it possible to merge it to master or would the change be too important ?

Btw thank you for your work on this very simple and efficient tool !

[hartym]

If you can work on a backport to master (without new features, just bug fix), I'm ok to merge it, although develop should be promoted to 0.7 quite soon (~1month) so I'm not certain it's worth the hassle.

Is it an option to use SetFields just before your CSV writer so that the get_input_fields() result contains the right thing ?

[yanjost]

I see SetFields sets the "output fields", but CsvWriter uses the "input fields", did I miss something ?

[hartym]

Well, it sets its own output fields, if you pipe Ainto SetFields(...) into CsvWriter(...), then SetFields' output fields are also CsvWriter's input fields.

[yanjost]

Nice ! I didn't understand that by reading the documentation, my bad !

So I think that a workaround being provided and the bugfix coming soon, this issue can be closed

[hartym]

(I'm making sure that what I said actually works, will close the issue then)

[hartym]

Actually there is an issue in 0.7 related to that and that was not known to me, I did a quick fix in 84749be2c80ae67f2833010673e6217d0fffb97c ; now testing 0.6.

[hartym]

I confirm the bug did not exist in stable version.

Here is my minimal test case for 0.6:

import bonobo

def get_graph():
    graph = bonobo.Graph()
    graph.add_chain(
        [(1, 2), (3, 4), (5, 6)],
        bonobo.SetFields(["x", "y"]),
        bonobo.CsvWriter("output.csv"),
    )
    return graph

if __name__ == "__main__":
    with bonobo.parse_args() as options:
        bonobo.run(get_graph())

Or the same thing using the new 0.7 syntax (old one still works of course):

import bonobo

def get_graph():
    graph = bonobo.Graph()
    graph >> [(1, 2), (3, 4), (5, 6)] >> bonobo.SetFields(
        ["x", "y"]
    ) >> bonobo.CsvWriter("output.csv")
    return graph

if __name__ == "__main__":
    with bonobo.parse_args() as options:
        bonobo.run(get_graph())