Hi Will,

I hope you're feeling better today.  I was playing with the mapid tool
this morning, and liked having it at my fingertips to use.  Below are
some comments/mods/questions.  Of these (4) is the most important for
now.

Thanks,
Mike

----------

1)	The mapping table should be loaded at most once, which results in a
	dramatic speedup when mapid() is called many times in a single process.

2)	Not being able to find an ID in the table is an error condition that
    should be reported to stderr, not stdout.

3)	Please find at

			/xchip/gdac_data/freezes/PANCAN12/my_mapid.py

	an updated implementation (version 0.2.5) containining (1) and (2).
   
4) Also find at

		/xchip/gdac_data/freezes/PANCAN12/unknown-ids.tsv

	a list of TCGA barcodes which it could not map to UUIDs.  I was going
	to ask you to why that's the case, but then Dan came by and suggested
	that it's the MAF ids we fabricate on ingest ... so, since MAF samples
	must by definition have a valid ID, we somehow need to discontinue
	fabricating fake ones (if possible) in favor of re-using the real IDs
	that already exist when we inject the MAFs into our GDAC data tree.