test(c/driver/postgresql): add test for composite type behavior

[amoeba]

No description provided.

[amoeba]

Still working on this, currently this fails with,

/Users/bryce/src/apache/arrow-adbc/c/driver/postgresql/postgresql_test.cc:1143: Failure
Expected equality of these values:
  reader.schema->children[0]->format
    Which is: "z"
  "i"

[lidavidm]

Hmm, maybe it isn't actually supported then...Or maybe you have to reconnect after creating the type to reload the type mapping

[amoeba]

I added what I think was a mid-test reconnect in 3e2eb6e. The new test still fails,

adbc ❯ driver/postgresql/adbc-driver-postgresql-test --gtest_filter="*Composite*"                                                                   (adbc)
adbc-driver-postgresql-test(30459,0x204bd5f00) malloc: nano zone abandoned due to inability to reserve vm space.
Running main() from /Users/runner/miniforge3/conda-bld/gtest-split_1748319995326/work/googletest/src/gtest_main.cc
Note: Google Test filter = *Composite*
[==========] Running 1 test from 1 test suite.
[----------] Global test environment set-up.
[----------] 1 test from PostgresStatementTest
[ RUN      ] PostgresStatementTest.PostgresCompositeTest
/Users/bryce/src/apache/arrow-adbc/c/driver/postgresql/postgresql_test.cc:1893: Failure
Expected equality of these values:
  reader.schema->children[0]->format
    Which is: "z"
  "i"

[  FAILED  ] PostgresStatementTest.PostgresCompositeTest (103 ms)
[----------] 1 test from PostgresStatementTest (103 ms total)

[----------] Global test environment tear-down
[==========] 1 test from 1 test suite ran. (103 ms total)
[  PASSED  ] 0 tests.
[  FAILED  ] 1 test, listed below:
[  FAILED  ] PostgresStatementTest.PostgresCompositeTest

 1 FAILED TEST

[amoeba]

Hi @lidavidm could you take a look at my test to see if that looks like I'm testing it correctly? If so, I'm not sure how hard adding support would be but I could at least update #3195 to document the current status.

[lidavidm]

Sorry this is on my backlog to poke

[amoeba]

No problem at all. Thanks.

[lidavidm]

I gave it a test in Python, creating the type beforehand in psql, and it worked. I think the AdbcDatabase needs to be recreated as that's what caches the type info. I think elsewhere someone requested a way to drop and recreate the type cache and maybe we should consider that

[lidavidm]

In [7]: cur.execute("SELECT * FROM asdf")

In [8]: cur.fetchallarrow()
Out[8]: 
pyarrow.Table
a: struct<x: int32, y: string>
  child 0, x: int32
  child 1, y: string
----
a: [
  -- is_valid: all not null
  -- child 0 type: int32
[1]
  -- child 1 type: string
["foo"]]

[amoeba]

Very interesting, thanks for figuring it out.

I pushed a fixed test which uses a test helper you may or may not like. I thought about making a new fixture but just adding a ResetTest method on StatementTest was near at hand. Happy to rework this.

[amoeba]

Rebased.

[lidavidm]

Hmm, ASAN is unhappy about something.

[amoeba]

I'll take a look later this week.

[lidavidm]

@amoeba do you think you want to do this for the next release?

[amoeba]

Hey @lidavidm, I won't be able to get this in for the release. I'll try for the next one.