Inconsistency in return type format on Intel and M1 Mac

[dpage]

I recently ran into an issue when working on pgAdmin where a scalar query result was returned as a string on Intel and bytes on M1. Here's a simple test case:

M1:

Python 3.10.11 (v3.10.11:7d4cc5aa85, Apr  4 2023, 19:05:19) [Clang 13.0.0 (clang-1300.0.29.30)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>> import psycopg
>>> with psycopg.connect("dbname=postgres user=dpage") as conn:
...     with conn.cursor() as cur:
...             cur.execute("SELECT version()")
...             cur.fetchone()
... 
<psycopg.Cursor [TUPLES_OK] [INTRANS] (user=dpage database=postgres) at 0x103a72610>
(b'PostgreSQL 15.1 on aarch64-apple-darwin22.1.0, compiled by Apple clang version 14.0.0 (clang-1400.0.29.202), 64-bit',)

Intel:

Python 3.10.4 (v3.10.4:9d38120e33, Mar 23 2022, 17:29:05) [Clang 13.0.0 (clang-1300.0.29.30)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>> import psycopg
>>> with psycopg.connect("dbname=postgres user=dpage") as conn:
...     with conn.cursor() as cur:
...             cur.execute("SELECT version()")
...             cur.fetchone()
...
<psycopg.Cursor [TUPLES_OK] [INTRANS] (user=dpage database=postgres) at 0x10e6c7530>
('PostgreSQL 15.2 (Homebrew) on x86_64-apple-darwin22.3.0, compiled by Apple clang version 14.0.0 (clang-1400.0.29.202), 64-bit',)

It seems to me that this should be consistent.

[dlax]

Can you also share conn.info.encoding in both cases?

[dvarrazzo]

Most definitely it should.

The only way psycopg returns bytes (I mean designed, as opposite as bugs) is that the client encoding of the connection is SQL_ASCII.

import psycopg
conn = psycopg.connect()
conn.execute("show client_encoding").fetchone()[0]
# 'UTF8'
conn.execute("set client_encoding to sql_ascii")
# <psycopg.Cursor [COMMAND_OK] [INTRANS] (database=piro) at 0x7fe76707e6c0>
conn.execute("show client_encoding").fetchone()[0]
# b'SQL_ASCII'

What does an execute("SHOW client_encoding").fetchone() from those connections show?

[dpage]

You've hit the nail on the head (as usual :-) ). My test server on Intel was initialised in UTF-8, but (for some unknown reason) SQL_ASCII on the M1 box.

Returning bytes for SQL_ASCII seems perfectly reasonable; I'll fix our code to handle that case.

Thanks!

[dvarrazzo]

No worries :) For completeness, strings adaptation rules are explained here.