/ref/taxonomy in HDF has tax_ids and parent IDs as objects. Should be integers

[jgolob]

The taxonomy cached in the HDF has the tax_ids as objects rather than integers. Same with parent IDs. Ideally both would be integers, to facilitate merges later.

[sminot]

Ah yes, the age-old debate of tax IDs as strings vs. integers! In this case I come down on the side of 'always strings' for three reasons:

• Joining on a string is as easy as joining on an integer
• An integer column with missing values gets coerced to a float (by Pandas), while strings are safe
• Only the current NCBI Taxonomy scheme happens to use integers for all tax IDs -- in the future we might end up adding tax IDs containing letters as well as numbers, and it's good to be future compatible

Looking forward to hearing your thoughts, @jgolob

[jgolob]

The case here for strings makes sense. I may adjust this instead to be then, /ref/taxonomy and /annot/gene/tax both use strings or integers for the tax_id column. Currently, /ref/taxonomy uses strings and /annot/gene/tax uses integers. Perhaps swap the latter to be strings?

[sminot]

Great idea! I'll add that now

[sminot]

Closed by a9751f5