ER_TOO_LONG_IDENT error when using pre-aggreations in MySQL

[gizmodus]

Describe the bug
As discussed with @paveltiunov: I get the following error when using pre-aggregations with MySQL:
ER_TOO_LONG_IDENT: Identifier name 'xxxxxxxx_xxxxxxxxxxxxxxxxx_category_and_date_grxbl5i2_qpwevmo0_1555596567123' is too long

To Reproduce
Define a pre-aggregation in the cube.js schema with a long name.

Version:
0.4.5

[paveltiunov]

Hey Reto! Thanks for posting this! MySQL is very restrictive in terms of table name length. First of all I think we can replace integer timestamp with string version.
Here's the place that parses timestamp:
https://github.com/statsbotco/cube.js/blob/50f1bbb7aac34ae7c4fd82a86ecee8506b17f4ea/packages/cubejs-query-orchestrator/orchestrator/PreAggregations.js#L40
Here's the place that generates table names:
https://github.com/statsbotco/cube.js/blob/50f1bbb7aac34ae7c4fd82a86ecee8506b17f4ea/packages/cubejs-query-orchestrator/orchestrator/PreAggregations.js#L255

[gizmodus]

@paveltiunov What do you mean with string version of the timestamp? Like date + time, e.g. "201904231718". This would save 2 characters 😃

[paveltiunov]

@gizmodus Hey Reto! I mean we can use for example the same base32 encoding we use for hash.

[beefs]

What about having a table to store the mapping of the tables?

[paveltiunov]

@beefs Hey Sunny! Great question! Cube.js was designed to minimize requirements for target DB in terms of pre-aggregations and one of design decision was not to store metadata in DB itself due to many big data backends doesn't support consistent transactional inserts/updates.

[lihan]

This issue happens on Postgres too, the table names are truncated and will not work.