Support viewing snapshot id history for table in branches/tags

[ajantha-bhat]

For timetravel queries snapshot id is required.
for main reference it works.
https://iceberg.apache.org/spark-queries/#:~:text=SELECT%20*%20FROM%20prod.db.table.history

But currently cannot see the snapshot id for branches in spark-sql extensions

Nessi CLI display

ajantha@Ajanthas-MacBook-Pro nessie % nessie contents -l -r main
ICEBERG_TABLE:
	db6.t2
	db5.t2

ajantha@Ajanthas-MacBook-Pro nessie % nessie contents -l -r b1  
ICEBERG_TABLE:
	db5.t2

spark-sql display

spark-sql> use nessie.db5;
Time taken: 0.027 seconds

spark-sql> select * from `nessie.db5.t2@b1`;
12
12
12
Time taken: 0.352 seconds, Fetched 3 row(s)

spark-sql> select * from `nessie.db5.t2.history@b1`;
Error in query: Table or view not found: `nessie.db5.t2.history@b1`; line 1 pos 14;
'Project [*]
+- 'UnresolvedRelation [nessie.db5.t2.history@b1]

spark-sql> select * from nessie.db5.t2.history;
2021-10-08 14:18:43.993	8461333400033237976	NULL	true
Time taken: 0.237 seconds, Fetched 1 row(s)

spark-sql> select * from nessie.db5.t2;
12
Time taken: 0.102 seconds, Fetched 1 row(s)

spark-sql> 

[omarsmak]

I am bit confused here, b1.history isn't Nessie branch, isn't? Shouldn't be something like this nessie.db5.t2.history@b1 to select history tables at a specific branch?

[ajantha-bhat]

@omarsmak : sorry for causing confusion, I have updated it. Still the problem exist with proper name.

[ajantha-bhat]

I see that CLI has content command to get the contents, but it shows content id and metadata pointer. I need snapshot id for the commits in the branch.

P.S: Also couldn't use contents CLI to get some info. waiting for document update with examples (#2277)

[snazy]

You could use table@ref_name or table@ref_name#commit_hash (latter only w/ the new Iceberg-Nessie).
Why do you need the snapshot-ID?

[ajantha-bhat]

If table@ref_name#commit_hash is supported then it is enough to do time travel queries within a branch. But it was not working when I tested. So, I thought we can expose snapshot id instead.

But Once the new changes are merged to support table@ref_name#commit_hash. I will close the issue by mentioning the PR. (or I should keep this issue in iceberg instead of nessie ?)

[snazy]

It's part of the PR/branch for Iceberg (which now uses the Nessie 0.10.1 release).

[ajantha-bhat]

Fixed in apache/iceberg#3257