SQL Lab does not show table names in the dropdown box

[rimolive]

A clear and concise description of what the bug is.
While selecting Database, schema, and table to inspect the table structure, I noticed that the table dropdown box only shows the schema name, not the table name. If I try selecting any of the options, I get an An error occurred while fetching table metadata error.

Expected results

I can see the columns that belong to the table

Actual results

An An error occurred while fetching table metadata error message appear

Screenshots

https://imgur.com/tfSfvyg

How to reproduce the bug

1. Go to 'SQL Lab > SQL Editor'
2. Click on any Database in the dropdown box
3. After refreshing the schema dropdown box, select any schema with tables
4. Check the tables dropdown box. It should appear only the schema name, not the tables name

Environment

• superset version: 0.37.1
• python version: Python 3.6.12
• node.js version: N/A
• npm version: N/A

Checklist

Make sure these boxes are checked before submitting your issue - thank you!

• I have checked the superset logs for python stacktraces and included it here as text if there are any.
• I have reproduced the issue with at least the latest released version of superset.
• I have checked the issue tracker for the same issue and I haven't found one similar.

Additional context

N/A

[issue-label-bot]

Issue-Label Bot is automatically applying the label #bug to this issue, with a confidence of 0.96. Please mark this comment with 👍 or 👎 to give our bot feedback!

Links: app homepage, dashboard and code for this bot.

[rimolive]

[mistercrunch]

What database engine is this?

[rimolive]

@mistercrunch The engine in question is Hive.

[bkyryliuk]

Hm, can't repro on the latest master for both hive & presto.
@rimolive maybe provide more info:

1. what hive version
2. what pyhive version

[rimolive]

pyhive version is 0.6.1. As for hive version it's a bit more complicated. :)

Actually, we use Spark to enable a Hiveserver2 special implementation to run queries against it. Currently Spark 2.4.3 is being used but we have plans to use Spark 3.0.0.

[fan-code]

pyhive version is 0.6.1. As for hive version it's a bit more complicated. :)

Actually, we use Spark to enable a Hiveserver2 special implementation to run queries against it. Currently Spark 2.4.3 is being used but we have plans to use Spark 3.0.0.

I also find this question when I use superset to connect to SparkThrift server.
My connection url is hive://ip:port/default?auth=KERBEROS&kerberos_service_name=hive .
Spark version is 2.4.0, How can I solve this problem?

[fan-code]

Hm, can't repro on the latest master for both hive & presto.
@rimolive maybe provide more info:

1. what hive version
2. what pyhive version

I also have the similar problem. hive connect is ok, but when connect to the sparkThrift，table scheme cann't appear table name, only schema name and the schema_name's number is euqal to talbe_number. eg. ods database has 100 tables, now ,you will see 100 ods in table schema in superset's sqllab
url is hive://ip:10001/default?auth=KERBEROS&kerberos_service_name=hive
hive version is 2.1.1+cdh6.2.0
pyhive version is 0.6.3

[rimolive]

Tracking requests from Superset UI, I see that when you select a Database the following URL is requested:

http://<superset-hostname><:superset_port>/superset/schemas/1/false/

Which returns a JSON with the following contents and populates the Schema selector:

{"schemas": ["main","examples"]} # This is just an example JSON output returned

When you select any schema, the following URL is requested:

http://<superset-hostname><:superset_port>/superset/tables/1/main/undefined/false/

This undefined in the URL clearly indicates that the Schema selector is not building the component correctly.

I'm still debugging the code and if I have a fix I'll cook a PR for it. Stay tuned!

Edited: hmm, so undefined seems acceptable but anyways the code that responds for that URL seems the issue. Still investigating.

[rimolive]

It is a bug in pyhive:

https://github.com/dropbox/PyHive/blob/master/pyhive/sqlalchemy_hive.py#L364

That line should not return row[0] since the result will be the schema name, not the table name. I'm going to propose a fix in pyhive.

Meanwhile, can this issue keep opened until pyhive team accept and fixes in latest releases?

Edit: Problem is even worse. See dropbox/PyHive#146
I'll see how to workaround this.

[junlincc]

@rimolive Ricardo, any updates?

[rimolive]

Unfortunately, I did not have an answer from Spark community but definitely this is a compatibility issue between Spark HiveServer2 implementation and Hadoop one. For now, I have no idea how this can be fixed.