Fixes #2164: Adds support for PostgreSQL 18

[sanchayanghosh]

Fixes #2164

1. In PostgreSQL 18, TupleDescAttr is now used to access ScanTuple->tts_tupleDescriptor->attrs since attrs is now replaced by compact_attrs to save on memory in PostgreSQL 18. In PostgreSQL 16, 17 and 18 we have a function TupleDescAttr which allows to acess pg_attrs
2. In PostgreSQL palloc0fast is now merged into palloc0.
3. Few funcitons now reuiqre executor/executor.h and are no longer present in the other includes.

[sanchayanghosh]

• Added sql tests for path deletion operations, ensuring persistence in temporary memory.

• Added a delete test to ensure that proper error message is shown.

• Investigated and found that SET commands are already tested for vertices and edges inside paths.

[sanchayanghosh]

@jrgemignani @MuhammadTahaNaveed

I didn't get the option to request reviewers. I would greatly appreciate if you could review my code? Always happy to incorporate any improvements.

Thanks!

[jrgemignani]

@sanchayanghosh Sorry for the delay. This would need to be added to a branch specifically for PG18 support.

[jrgemignani]

@sanchayanghosh I have created PG18_prepare for such work. It will need to be adjusted a bit before it actually is ready, but it would be where you would apply this PR.

[jrgemignani]

@sanchayanghosh I have created PG18_prepare for such work. It will need to be adjusted a bit before it actually is ready, but it would be where you would apply this PR.

@sanchayanghosh could you apply this against PG18_prepare ?

[MuhammadTahaNaveed]

@sanchayanghosh ty for the pr and sorry for the delay. I have changed the base branch to PG18_prepare. Since we maintain a separate branch for every PostgreSQL version, if you can remove the version guard macros and just keep the changes required for PG18 compatibility?

[sanchayanghosh]

I removed the macro checks. Thanks to you both for reviews. I am sorry I was in middle of some other projects I couldn't check last month and got free this week onwards.

[jrgemignani]

I removed the macro checks. Thanks to you both for reviews. I am sorry I was in middle of some other projects I couldn't check last month and got free this week onwards.

No worries. This is a volunteer project and we all have other things to do.

[jrgemignani]

@jrgemignani @MuhammadTahaNaveed

I didn't get the option to request reviewers. I would greatly appreciate if you could review my code? Always happy to incorporate any improvements.

Thanks!

I added Taha and myself as reviewers.

[MuhammadTahaNaveed]

@sanchayanghosh Regression tests fail and there's a crash in cypher_match regression test. Maybe its due to reason mentioned by @harudini here #2164 (comment).

ok 1         - scan                                      166 ms
ok 2         - graphid                                     9 ms
ok 3         - agtype                                     93 ms
ok 4         - agtype_hash_cmp                             7 ms
ok 5         - catalog                                    73 ms
ok 6         - cypher                                     37 ms
not ok 7     - expr                                      668 ms
ok 8         - cypher_create                             105 ms
not ok 9     - cypher_match                              852 ms
# (test process exited with exit code 2)
not ok 10    - cypher_unwind                              21 ms
# (test process exited with exit code 2)
not ok 11    - cypher_set                                  2 ms
# (test process exited with exit code 2)
not ok 12    - cypher_remove                               2 ms
# (test process exited with exit code 2)
not ok 13    - cypher_delete                               3 ms
# (test process exited with exit code 2)
not ok 14    - cypher_with                                 2 ms
# (test process exited with exit code 2)
not ok 15    - cypher_vle                                  3 ms
# (test process exited with exit code 2)
not ok 16    - cypher_union                                2 ms
# (test process exited with exit code 2)
not ok 17    - cypher_call                                 2 ms
# (test process exited with exit code 2)
not ok 18    - cypher_merge                                3 ms
# (test process exited with exit code 2)
not ok 19    - cypher_subquery                             2 ms
# (test process exited with exit code 2)
not ok 20    - age_global_graph                            2 ms
# (test process exited with exit code 2)
not ok 21    - age_load                                    2 ms
# (test process exited with exit code 2)
not ok 22    - index                                       2 ms
# (test process exited with exit code 2)
not ok 23    - analyze                                     2 ms
# (test process exited with exit code 2)
not ok 24    - graph_generation                            2 ms
# (test process exited with exit code 2)
not ok 25    - name_validation                             2 ms
# (test process exited with exit code 2)
not ok 26    - jsonb_operators                             3 ms
# (test process exited with exit code 2)
not ok 27    - list_comprehension                          2 ms
# (test process exited with exit code 2)
ok 28        - map_projection                             37 ms
ok 29        - cypher_path_delete                         43 ms
not ok 30    - drop                                       57 ms
# (test process exited with exit code 2)
1..30
# 21 of 30 tests failed.

[jrgemignani]

@sanchayanghosh Regression tests fail and there's a crash in cypher_match regression test.

Same for me. How did you build this prior to submitting it? Did it work for you?

[jrgemignani]

@sanchayanghosh @MuhammadTahaNaveed Is this necessary to add to this PR? Could, should, it be its own PR?

Added a delete test to ensure that proper error message is shown.

[jrgemignani]

@sanchayanghosh Any updates?

[sanchayanghosh]

Sorry for delay. Had a big project that took my time

I will split the tests into its own PR and illoom into the crash once more

[jrgemignani]

Sorry for delay. Had a big project that took my time

I will split the tests into its own PR and illoom into the crash once more

@sanchayanghosh No worries.

We did move this to an updated PG18 branch. You will need to rebase your work off of it.

[optionals]

-- load 'age';
-- SET search_path = ag_catalog, "$user", public;

select * from create_graph('knowledge_graph');

SELECT
    *
FROM
    cypher ('knowledge_graph', $$
                        MERGE (n:TOPIC {id: '3tbr5xae-topic-ft1a', kg_id: '3tbr5xae'})
                        RETURN n$$) AS (n agtype);


PREPARE age_merge_topic (agtype) AS
                    SELECT * FROM cypher('knowledge_graph', $$

                        MATCH (n:TOPIC {id: $id, kg_id: $kg_id})
                        SET n.name = $name
                        RETURN n
                    $$, $1) AS (n agtype);

-- crash in postgres 18.1
EXECUTE age_merge_topic('{"id": "3tbr5xae-topic-ft1a","kg_id": "3tbr5xae", "name": "12"}'::agtype);


-- fine
PREPARE cypher_stored_procedure(agtype) AS
SELECT *
FROM cypher('knowledge_graph', $$
    MATCH (v:Person) 
    WHERE v.name = $name
    RETURN v
$$, $1)
AS (v agtype);

EXECUTE cypher_stored_procedure('{"name": "Tobias"}');

The above statements will cause the database to crash: Segmentation fault: 11, currently only found in the set operation.

But this statement runs normally when using the Docker version of apache/age (Debian 17.7-3.pgdg13+, age 1.6.0).

macbook pro with with m4 , macOS 26.1
posgtgresql 18.1 from Homebrew

[jrgemignani]

@optionals Apache AGE PG18 isn't working with PostgreSQL 18 yet.

[optionals]

@optionals Apache AGE PG18 isn't working with PostgreSQL 18 yet.

I used the code from this PR, Isn't this PR just for adapting to PostgreSQL 18?

I tried the branch "sanchayanghosh:feature/2164/provide-support-for-postgresql-16-18," and also tried merging it into the latest master code: 571c19

[jrgemignani]

@optionals Apache AGE PG18 isn't working with PostgreSQL 18 yet.

I used the code from this PR, Isn't this PR just for adapting to PostgreSQL 18?

I tried the branch "sanchayanghosh:feature/2164/provide-support-for-postgresql-16-18," and also tried merging it into the latest master code: 571c19

PG18 is still a test branch and this particular PR currently does not allow PG18 to build with Apache AGE.

[optionals]

@optionals Apache AGE PG18 isn't working with PostgreSQL 18 yet.

I used the code from this PR, Isn't this PR just for adapting to PostgreSQL 18?
I tried the branch "sanchayanghosh:feature/2164/provide-support-for-postgresql-16-18," and also tried merging it into the latest master code: 571c19

PG18 is still a test branch and this particular PR currently does not allow PG18 to build with Apache AGE.

I understand. This issue was discovered in a local development environment. The comments above are simply to report that this PR has this problem. Before merging, it should be verified whether the issue indeed exists, and if so, a targeted fix will be necessary.

In addition, I used this approach because the issue described in #1722 currently has no suitable solution. If a workaround or an updated fix becomes available, it would be greatly appreciated. After multiple attempts locally, I have not found any alternative besides fully manual SQL construction or using PREPARE as a workaround.

[jrgemignani]

@optionals Apache AGE PG18 isn't working with PostgreSQL 18 yet.

I used the code from this PR, Isn't this PR just for adapting to PostgreSQL 18?
I tried the branch "sanchayanghosh:feature/2164/provide-support-for-postgresql-16-18," and also tried merging it into the latest master code: 571c19

PG18 is still a test branch and this particular PR currently does not allow PG18 to build with Apache AGE.

I understand. This issue was discovered in a local development environment. The comments above are simply to report that this PR has this problem. Before merging, it should be verified whether the issue indeed exists, and if so, a targeted fix will be necessary.

In addition, I used this approach because the issue described in #1722 currently has no suitable solution. If a workaround or an updated fix becomes available, it would be greatly appreciated. After multiple attempts locally, I have not found any alternative besides fully manual SQL construction or using PREPARE as a workaround.

@optionals -

You may want to try PR #2251. It is another PR to bring PG18 into the fold and it looks good so far. Your test could be valuable there as well.

[optionals]

@optionals Apache AGE PG18 isn't working with PostgreSQL 18 yet.

I used the code from this PR, Isn't this PR just for adapting to PostgreSQL 18?
I tried the branch "sanchayanghosh:feature/2164/provide-support-for-postgresql-16-18," and also tried merging it into the latest master code: 571c19

PG18 is still a test branch and this particular PR currently does not allow PG18 to build with Apache AGE.

I understand. This issue was discovered in a local development environment. The comments above are simply to report that this PR has this problem. Before merging, it should be verified whether the issue indeed exists, and if so, a targeted fix will be necessary.
In addition, I used this approach because the issue described in #1722 currently has no suitable solution. If a workaround or an updated fix becomes available, it would be greatly appreciated. After multiple attempts locally, I have not found any alternative besides fully manual SQL construction or using PREPARE as a workaround.

@optionals -

You may want to try PR #2251. It is another PR to bring PG18 into the fold and it looks good so far. Your test could be valuable there as well.

I tried #2251, the above statement works fine, thanks. (only the above scenarios have been tested.)

[sanchayanghosh]

The other PR solved the problem and that solution was more complete than mine. I will close this PR until further issue.

[jrgemignani]

@sanchayanghosh Regardless, I want to thank you for your efforts!