zeekayandHanzo Dev 752f11a7c4 Count the rows a typed query reads, on PostgreSQL too ...

Typed[T].Count swapped the projection for COUNT(*) and kept every other
clause. SQLite accepted the ORDER BY left beside the aggregate; PostgreSQL
refuses it ("must appear in the GROUP BY clause"), so every ordered Count
failed on hanzo-sql alone. A grouped query answered one group's count, and
LIMIT and DISTINCT were ignored on both. Count now runs the query itself as
a subquery of SELECT COUNT(*), with the order dropped.

It was found by running the relational tests on PostgreSQL, which they now
do on every run: internal/pgtest starts a throwaway server from the
binaries on the machine, on a Unix socket in a temporary directory, and a
machine without one says so by name.

replicated/go.mod was untidy against the root module, which failed its
build gate on main.

Co-authored-by: Hanzo Dev <[email protected]>

2026-09-19 10:05:50 -07:00

..

go.mod

Count the rows a typed query reads, on PostgreSQL too

2026-09-19 10:05:50 -07:00

go.sum

Count the rows a typed query reads, on PostgreSQL too

2026-09-19 10:05:50 -07:00

replicated_test.go

replicated: build against the tree, not the last published orm

2026-07-27 20:25:39 -07:00

replicated.go

db: fold namespace case, and document the seam hanzoai/namespace takes

2026-07-28 09:45:55 -07:00