biocore / songbird

Vanilla regression methods for microbiome differential abundance analysis
BSD 3-Clause "New" or "Revised" License
54 stars 25 forks source link

Make error message on invalid training column values more clear #102

Open fedarko opened 4 years ago

fedarko commented 4 years ago

@vjcantu and I ran into this today -- if values in this column are labelled e.g. train and test, then 0 Train samples are identified (see line 189 below) and then later on Tensorflow fails with an obscure error that num_classes should be positive, got 0.

https://github.com/biocore/songbird/blob/22ec2b550a5a0e5f3d4ae134dbf0a963616e64ad/songbird/util.py#L178-L197

Even if we'd prefer to leave this as case sensitive, it might be good to add a note to the README/FAQs explaining where this error comes from and/or that the training column stuff is case sensitive.

mortonjt commented 4 years ago

👍 on the README update.

Note that we already have this explicit in the README and the CLI documentation (see this line). So maybe screenshots or notes on the exact casing is necessary.