FileIO shard numbering would be more user-friendly if it were 1-based instead of 0-based

[iht]

What happened?

When fileio writes several shards, the last shard will have as name "00123-of-00124" which may look like incomplete processing.

The first shard should be "00001-of-XXXXX" and not "00000-of-XXXX", and the last shard should be "00124-of-00-124" (in this case for an example with 124 shards).

Issue Priority

Priority: 3

Issue Component

Component: io-py-files

[kennknowles]

Acknowledged and I do agree. I don't think this should cause any backwards-incompatibility to change but @johnjcasey and @reuvenlax would know better, regarding pipeline update and how it might interact here. Obviously this is a classic bikeshed.

[kennknowles]

This occurs in Python and Java and also in internal Dataflow code separate from both of them, I expect.