Question: are there significant downsides to increasing IDSIZE?

Hi,

I just made a fork of DReichLab/EIG where I:

• increased IDSIZE in include/admutils.h from 40 to 100
• increased the s1 and s2 char arrays in src/h2d.c from length 50 to 120

I did this because the population labels in my ind/fam files are often very descriptive, in the form of e.g. <ETHNICITY>_<REGION1>_<REGION2>[_<DIFFERENT_ETHNICITY>Inferred], and they sometimes go over the old limit and make smartpca etc error out.

After re-makeing everything and running the new smartpca binary, I successfully avoid the idnames too long error, and the evec output from the toy data at POPGEN/example and from two sets of real data all look fine.

I understand that you wouldn't want to set IDSIZE to something stupid like a million, but in every plausible use case I can imagine I can't see my 39 to 99 change significantly increasing the memory footprint over the baseline. Are there other potential downsides to my button-mashing that I'm too short-sighted to conceive of?

Review include/admutils.h and src/h2d.c, then inspect how IDSIZE, s1, and s2 are used throughout the C sources. Rebuild with make and run smartpca on POPGEN/example plus representative ind/fam inputs; done means identifying any compatibility, memory, or output risks and recording a maintainer decision.