The Rows interface in pgx, like its ancestor in database/sql, is easy to
accidentally misuse in a way that can cause apps to misinterpret
database or connection errors as successful queries with empty or
truncated result-sets.
Update the docs to emphasise the need to call rows.Err() after
rows.Next() returns false, and direct users of the interface to the v5
API helpers that make writing correct code easier.
The docs on Conn.Query() already call this out, so only a small change
is needed to warn users and point them at the details on Query()
Per details in #1707
Previously it was checking TotalConns but that includes ConstructingConns.
Instead it should directly check IdleConns so the next Acquire takes one of
those and doesn't make a 3rd connection. The check against the context was
also wrong which prevented this from timing out after 2 minutes.
This also fixes a bug where NewConnsCount was not correctly counting
connections created by Acquire directly.
Fixes#1690
It was previously possible for a connection to be created while the
background health check was running. The health check could create
connection(s) in excess of the maximum pool size in this case.
https://github.com/jackc/pgx/issues/1660
connect_timeout given in conn string was not obeyed if sslmode is not specified (default is prefer) or equals sslmode=allow|prefer. It took twice the amount of time specified by connect_timeout in conn string. While this behavior is correct if multi-host is provided in conn string, it doesn't look correct in case of single host. This behavior was also not matching with libpq.
The root cause was to implement sslmode=allow|prefer conn are tried twice. First with TLSConfig and if that doesn't work then without TLSConfig. The fix for this issue now uses the same context if same host is being tried out. This change won't affect the existing multi-host behavior.
This PR goal is to close issue [jackc/pgx/issues/1672](https://github.com/jackc/pgx/issues/1672)
The context timeouts for tests are designed to give a better error
message when something hangs rather than the test just timing out.
Unfortunately, the potato CI frequently has some test or another
randomly take a long time. While the increased times are somewhat less
than optimal on a real computer, hopefully this will solve the
flickering CI.
Most of this is in conversion, and I assume it became unused with some
of the v5 changes and refactors to a codec-based approach.
There are likely a few more cleanups to be made, but these ones seemed
easy and safe to start with.
As recommended by go-staticcheck, but also might be a bit more efficient
for the compiler to implement, since we don't care about which slice of
bytes is greater than the other one.
This ensures the output of Encode can pass through Scan and produce
the same input. This found two two minor problems with the text
codec. These are not bugs: These situations do not happen when using
pgx with Postgres. However, I think it is worth fixing to ensure the
code is internally consistent.
The problems with the text codec are:
* It did not correctly distinguish between nil and empty. This is not
a problem with Postgres, since NULL values are marked separately,
but the binary codec distinguishes between them, so it seems like
the text codec should as well.
* It did not output spaces between keys. Postgres produces output in
this format, and the parser now only strictly parses the Postgres
format. This is not a bug, but seems like a good idea.
Craig Ringer
Jack Christensen
Alexey Palazhchenko
horpto
James Hartig
Rafi Shamim
Evan Jones
smaher-edb
Christoph Engelbert (noctarius)
Dan McGee
Adrian-Stefan Mares
Gerasimos (Makis) Maropoulos
Brandon Kauffman
Nicola Murino
Lev Zakharov