unable to read data from oracle catalog via trino

[vmsaipreeth]

We are using trino as our backend.

connection = ibis.trino.connect( trino_connection_parameters)

df_ibis = connection.sql(“””SQL STATEMENT WITH ORACLE CATALOG VIA TRINO”””)

This throws an error:

ParseError: expected 'EOF' at 0:7

Note: This works fine for hdfs and object storage catalogs

[gforsyth]

Hi @vmsaipreeth -- sorry to hear you are having trouble.

We will be in a much better position to help you with these issues if you can provide reproducible examples.

At the very least, we will require a representative example and the full traceback that is associated with that error.

[vmsaipreeth]

This is how we are trying to fetch data from an oracle catalog via trino

df_ibis = connection.sql(“”” SELECT col_a, col_b from ORACLE_CATALOG.SCHEMA.TABLE”””)

The table as many columns. However choosing only two columns for now.

DataTypes of the columns:

col_a decimal(15, 0) NOT NULL,
col_b varchar(7) NOT NULL

This throws an error while parsing the above SQL statement:

ParseError: expected 'EOF' at 0:7

Please note that connection is established successfully and we are able to fetch data for hdfs and object storage catalogs. Observing the above issue only for oracle catalogs.

[cpcloud]

I suspect this is a type parsing issue on the ibis side.

[Tay0z]

@cpcloud is there a way to resolve this issue using a different query method?

[gforsyth]

@Tay0z -- if the Oracle database is available, you can try using the new Oracle backend to connect to it directly. In any case, we'll attempt to fix the parsing issue.

[cpcloud]

This was a bug in parsing trino types. We hadn't implemented parsing for the all the types it supports.

Addressed in #6658

[vmsaipreeth]

Thanks @cpcloud for quick response. Highly appreciate it.
Btw, when can we expect this feature to roll out?

[cpcloud]

I think we're going to release in a week or two, stay tuned!

[vmsaipreeth]

Awesome! Will keep an eye. Thank you!

[vmsaipreeth]

Hi there!

So, we upgraded ibis to 6.1.0 and had chance to play around. Thanks for the prompt response and fix.

We are now able to read oracle catalogs via trino 👍🏼.

Below code which uses connection.sql() works well

connection = ibis.trino.connect( trino_connection_parameters)

df_ibis = connection.sql(“””SQL STATEMENT WITH ORACLE CATALOG VIA TRINO”””)

However, we noticed that the reading time as increased. Below are few metrics:

Time for reading Table A from object storage: 3.6 secs
Time for reading Table A from oracle catalog: 27.3 secs

Time for reading Table B from object storage: 1.8 secs
Time for reading Table B from oracle catalog: 16.3 secs

is there anyway that we can reduce the reading time for oracle catalogs that you would suggest?

Thanks in advance!

[cpcloud]

Let's discuss in an issue!