offers: add used_count to listoffers

[alexgrad42]

This adds a new 'used_count' field to the 'listoffers' RPC command output to see the exact number of times a BOLT12 offer has been paid by counting associated settled invoices.

Changelog-Added: add a new 'used_count' field to the 'listoffers' RPC command

Closes #9146

[nGoline]

Great addition. This fills a real gap and the implementation is straightforward, the core idea is solid and it's close to being mergeable.
A few things to fix before we can land this.

[nGoline]

You're passing offer_id instead of &id here. Since we're inside the else block (meaning offer_id was not provided), this would pass NULL to a function annotated with NON_NULL_ARGS(1,2), so every offer in the list-all path would get a count of 0 (or worse).
Should be &id.

[nGoline]

You should remove the ; between the function declaration and NON_NULL_ARGS.

Suggested change

	u64 wallet_offer_count_payments(struct wallet *w,
	const struct sha256 *offer_id);
	NON_NULL_ARGS(1,2);
	u64 wallet_offer_count_payments(struct wallet *w,
	const struct sha256 *offer_id)
	NON_NULL_ARGS(1,2);

[nGoline]

The hardcoded 1 is fragile and doesn't self-document intent. We have a enum constant for PAID.

Suggested change

	" AND state = 1;"));
	db_bind_sha256(stmt, offer_id);
	" AND state = ?;"));
	db_bind_sha256(stmt, offer_id);
	db_bind_int(stmt, invoice_status_in_db(PAID));

[nGoline]

The indentation here is inconsistent with a mix of spaces and tabs. Please align it with the surrounding code (tabs only).

[nGoline]

pre-v0.10.1 is reserved for fields that existed before versioning was introduced. This is a new field, so it needs the correct release tag. There's no time to include it in v26.06, so let's target the next one: v26.09.

[nGoline]

used_count is always computed and returned unconditionally, so it should be listed in the required array of the items object in this schema.

[nGoline]

This executes one SELECT COUNT(*) per offer, which means a node with 1000 offers will fire 1001 queries on every listoffers call. Consider computing all counts in a single query with GROUP BY local_offer_id, then looking up each offer's count from the result.

[nGoline]

Trailing whitespace: make check-source will catch this. Please remove the space after ));..

[nGoline]

You're adding this field between json_populate_offer() output and description. Consider moving it after description so the JSON output is consistent between versions.

[nGoline]

Let's drop the custom message string. The in check is redundant since the == 0 assertion on the next line would already produce a KeyError if the key were missing. Plain assert is the project convention here.

Suggested change

	assert 'used_count' in offers[0], "Error: used_count field is missing from listoffers response!"
	assert offers[0]['used_count'] == 0
	assert offers[0]['used_count'] == 0

[nGoline]

Please rebase onto master instead of merging it in. Merging master into your feature branch creates a noisy history and makes the diff harder to follow. A clean rebase keeps the commits linear and makes it easier to bisect if something goes wrong later

[nGoline]

Please rebase onto master instead of merging it in. Merging master into your feature branch creates a noisy history and makes the diff harder to follow. A clean rebase keeps the commits linear and makes it easier to bisect if something goes wrong later

Also, when addressing the review feedback, please use fixup! commits (or git commit --fixup=<sha>) rather than separate "fix" commits. Having a commit like "fix indentation" or "address review comments" in the final history makes it harder to bisect.