fix(postgresql): loosen requirements for getPrimaryKeyWhereClause

[vinhloc30796]

substreams_history will look like so:

substreams=> select * from substreams_history limit 1;
-[ RECORD 1 ]-----------------------------------------------------------------------------------------------------------------------------------------------------------------------
id         | 75756
op         | I
table_name | "public"."subtransaction"
pk         | {"block_number":"34830163","chain_id":"8453","trace_index":"ARRAY[]::bigint[]","transaction_hash":"0xd24deec488102fe7356efb5c034545f8d9c945b8881764c8b7f70a62a3c9d2a4"}
prev_value | 
block_num  | 34830163

[maoueh]

Why is this a good way of doing it, can you add a comment explaining the reasoning here, we should also log the error minimally in DEBUG mode if it's expected to log a lot of time.

[maoueh]

Same here, add comment as the why it's correct and log error.

[maoueh]

Is this backward compatible? The question is essentially, if there is already substreams_history entries`, are they going to be read correctly?

[GabrielCartier]

Hi @vinhloc30796 — thanks for this, and sorry for the long silence.

This repository is being deprecated (see #133). The SQL sink now lives in the main substreams CLI as substreams sink postgres / substreams sink clickhouse, shipped in substreams v1.20.2. All future work happens there, so this PR can't land here.

The code moved as-is, so the port is mostly a path rewrite: db_changes/db/dialect_postgres.go is now sink/sql/db_changes/db/dialect_postgres.go in streamingfast/substreams.

The typed primary-key / array-literal handling isn't in the CLI yet, and the underlying problem is still real — if you're up for reopening it against develop there (with the review feedback from this thread addressed), we'll pick it up. The branch conflicts here anyway, so rebasing onto the new tree is roughly the same effort. If you'd rather not, tell us and we'll port it ourselves with credit to you.

Closing this one out.