Files with comma in name are not properly escaped in output CSV

[subdavis]

Caused https://viame.kitware.com/girder#job/606200e1667a04d5d4f8682f to fail.

Here's an example of the output from a file called 20171027_,_214855.865,02,9443.png

# 1: Detection or Track-id,  2: Video or Image Identifier,  3: Unique Frame Identifier,  4-7: Img-bbox(TL_x,TL_y,BR_x,BR_y),  8: Detection or Length Confidence,  9: Target Length (0 or -1 if invalid),  10-11+: Repeated Species, Confidence Pairs or Attributes
# Written on: Tue Mar 30 22:09:20 2021   by: write_detected_object_set_viame_csv
0,20171027_,_214855.865,02,9443.png,0,636.762,71.0875,712.387,189.062,0.998693,0,generic_object_proposal,0.998693
1,20171027_,_214855.865,02,9443.png,0,756.25,15.125,968,192.088,0.998171,0,generic_object_proposal,0.998171
2,20171027_,_214855.865,02,9443.png,0,766.837,4.5375,968,40.8375,0.968831,0,generic_object_proposal,0.968831
3,20171027_,_214855.865,02,9443.png,0,896.913,355.438,968,425.012,0.96825,0,generic_object_proposal,0.96825
4,20171027_,_214855.865,02,9443.png,0,898.425,251.075,968,350.9,0.929115,0,generic_object_proposal,0.929115
5,20171027_,_214855.865,02,9443.png,0,739.612,624.663,766.837,685.163,0.211674,0,generic_object_proposal,0.211674

DIVE could technically handle this by doing a name transform for inputs, but then they wouldn't match the input file name, which isn't ideal.

[russelldj]

We've been writing some somewhat brittle csv writers on the kwiver side. I was wondering if there are any library resources to help at least clean up cases like this. My guess is the "right" way to handle this is use quoting and have your reader and writer respect the same spec for it.

[subdavis]

If you're using python, the regular csv module will do everything perfectly if you use the writer.write([array, of, column, values]) function.

[russelldj]

We haven't been but that seems like it might be a good option moving forward as the python support improves. Thanks.