Identify and document scalability benchmarks

[mortonjt]

Empress needs to be run against a huge tree (> 1 million tips)

[antgonza]

Just wondering if there are any updates on this issue; thank you.

[antgonza]

Installed the latest version of empress and ran it on one of the large trees generated in Qiita within a 2020.2 Qiime2 conda environment; the mapping file, feature-table and taxonomies from the moving pictures dataset - only one dataset.

Note that this is a tree was created over a year ago (we could generate even larger today), is the 100bp fragments insertion tree and is ~8.8M tips:

In [1]: from skbio import TreeNode
In [2]: tree = TreeNode.read('../insertion_tree.relabelled.tre')
In [3]: print(tree.count(tips=True))     
8830174

I generated the no-taxonomy, GG and Silva added empress qzv's to test, each takes ~3hrs to generate the qzv and it works just fine (no error messages). However, when I try to open them in https://view.qiime2.org/, the browser fails with:

and if I unzip the qzv and try to open the index.html or empress.html, I get:

Anyway, here are the testing files.

cc: @ElDeveloper

[kwcantrell]

@antgonza I'm looking into this

[fedarko]

Once we identify upper bounds for what sorts of data sizes Empress can comfortably visualize, we should document this clearly in the README so that e.g. users with billion-tip trees know that they probably want to consult another tool and/or a priest ._.