Thread

Re: Feature: psql - display current search_path in prompt

Florents Tselai <florents.tselai@gmail.com> — 2025-06-10T13:37:57Z

> On 10 Jun 2025, at 4:09 PM, Florents Tselai <florents.tselai@gmail.com> wrote:
> 
> 
> 
>> On 10 Jun 2025, at 3:51 PM, Jim Jones <jim.jones@uni-muenster.de> wrote:
>> 
>> Hi Florents
>> 
>> On 10.06.25 13:36, Florents Tselai wrote:
>>> 
>>> On Tue, Jun 10, 2025 at 2:08 AM Jelte Fennema-Nio <postgres@jeltef.nl
>>> <mailto:postgres@jeltef.nl>> wrote:
>>> 
>>>    On Mon, 9 Jun 2025 at 17:54, Florents Tselai
>>>    <florents.tselai@gmail.com <mailto:florents.tselai@gmail.com>> wrote:
>>>> Here’s a quick attempt that makes %S substitue for a search_path
>>>> Like
>>>>   \set PROMPT1 'user:%n search_path: %S'
>>> 
>>>    +   else
>>>    +       return PQuser(pset.db);
>>> 
>>>    That seems like a copy paste error. If we don't have data for it, we
>>>    should either use the empty string, or some string like "<unknown>".
>>> 
>>> 
>>> Opted for an empty string
>>>  
>>> 
>>>    Other than that, the patch looks good (but I haven't tested it yet).
>>> 
>> 
>> 
>> I've taken a quick look at the patch, and it seems to work as expected.
>> 
>> == PROMPT1 ==
>> 
>> postgres=# \set PROMPT1 '(search_path: %S) ;; '
>> (search_path: "$user", public) ;; SET search_path TO s1, public;
>> SET
>> (search_path: s1, public) ;; SET search_path TO s2, public;
>> SET
>> (search_path: s2, public) ;; RESET search_path;
>> RESET
>> (search_path: "$user", public) ;;
>> 
>> == PROMPT2 ==
>> 
>> postgres=# \set PROMPT2 '(search_path: %S) ** '
>> postgres=# SELECT
>> (search_path: "$user", public) ** ^C
>> postgres=# SET search_path TO s1, public;
>> SET
>> postgres=# SELECT
>> (search_path: s1, public) ** ^C
>> postgres=# SET search_path TO s2, public;
>> SET
>> postgres=# SELECT
>> (search_path: s2, public) ** ^C
>> postgres=# RESET search_path;
>> RESET
>> postgres=# SELECT
>> (search_path: "$user", public) **
>> 
>> 
>> == PROMPT3 ==
>> 
>> postgres=# \set PROMPT3 '(search_path: %S) ## '
>> postgres=# COPY t1 (a) FROM STDIN;
>> Enter data to be copied followed by a newline.
>> End with a backslash and a period on a line by itself, or an EOF signal.
>> (search_path: "$user", public) ## 1
>> (search_path: "$user", public) ## 2
>> (search_path: "$user", public) ## \.
>> COPY 2
>> postgres=# SET search_path TO s1, public;
>> SET
>> postgres=# COPY t1 (a) FROM STDIN;
>> Enter data to be copied followed by a newline.
>> End with a backslash and a period on a line by itself, or an EOF signal.
>> (search_path: s1, public) ## 42
>> (search_path: s1, public) ## 73
>> (search_path: s1, public) ## \.
>> COPY 2
>> postgres=# RESET search_path;
>> RESET
>> postgres=# COPY t1 (a) FROM STDIN;
>> Enter data to be copied followed by a newline.
>> End with a backslash and a period on a line by itself, or an EOF signal.
>> (search_path: "$user", public) ## 0
>> (search_path: "$user", public) ## 1
>> (search_path: "$user", public) ## \.
>> COPY 2
>> 
>> Documentation looks ok as well -- it aligns with the other entries in
>> the file.
> 
> Thanks for the review. 
> 
>> 
>>> 
>>> Btw - I haven't worked on bin/psql code; 
>>> aren't these auto tested?  
>> 
>> I also couldn't find any test related to psql's PROMPT*. Perhaps Jelte
>> knows more about it?
> 
> 
> Doesn’t look like it though; 
> e.g. this https://github.com/Florents-Tselai/postgres/commit/79fad725aa410d6c631d4ffe0f4120837f9b478c
> didn’t break anything - on Cirrus at least

EDIT: There are test under `src/psql/t` , not sure though how much coverage they have,
but most importantly how it’d look like for this case.