BUG: evaluate-busco brakes when input mags > ~1,300

[Sann5]

What's new

• the plotting library vega has a limit on the size of the plots that can be computed.
• a user was trying to plot something really big and therefore the evaluate-busco action was erroring out.
• this PR fixes that such that even when the dataset is really big a visualization is produced.
• even though we can avoid error in the action there is no going around the limit on the size of the plot. If one tries to plot more than ~1,300 MAGs the output will be an error image instead of the plot.
• as a workaround to be able to work with datasets containing more than ~1,300 MAGs, we added a search box function to the visualization such that users can display an arbitrary subset of the samples. As long as this subset is bellow the allowed limit then a plot will be produced.
• Closes BUG: evaluate-busco brakes when input mags > ~1,300 #140

Run it locally

1. Assuming you already have a local copy of q2-moshpit that is installed in your virtual environment.

cd q2-moshpit
gh pr checkout 142

2. Let's get you some data to play with:

cd <download here>
wget https://polybox.ethz.ch/index.php/s/slGyW1uQbyxmyb9/download -O mags.qza

3. Test it out!

qiime moshpit evaluate-busco --i-bins mags.qza --verbose --p-lineage-dataset bacteria_odb10 --p-cpu 6 --o-visualization mags.qzv 

qiime tools view mags.qzv 

Running the tests

pytest -W ignore -vv --pyargs q2_moshpit

[misialq]

Hey @Sann5, thanks for fixing that! I changed the code slightly to remove the dependency on the vegatransformer - with this present I would constantly get errors that some packages are missing in the environment. Besides, this was needed to display more than 5k rows, right? So I guess we don't need it anyway, unless you kept it there on purpose? In any case, I'm merging this one now and we can see later whether we should think about that transformer again.

[Sann5]

I'm not super sure @michal but think the transformer might be needed. I can't remember if the visualization broke earlier (with a smaller dataset) or rather if it was needed for the search box function to work correctly. But if you tried this out with a large dataset and it worked then we are good to go :).