Change foreign key constraint names (#38) #39
Conversation
This changes foreign key constraint names to be based on the table and columns of the foreign key, so that they are unique. This is a breaking change.
|
We will need to think about providing a migration guide as part of this PR. |
|
I wrote my thoughts about migration for this issue. As noted in the solution section for this issue, this change does not fix the issues with the |
This changes the foreign key constraint names created by automatic foreign key discovery to include all constraint columns instead of just the first one, so that they are unique. This is a breaking change.
|
We should also be careful about the length of these names, since we're already running into problems where the generated names overlap due to length limitations. |
|
I have run into constraint name length issues in my project as well. Since foreign key column names are prefixed with the table name when using Beam naming conventions, the column names can be pretty long. The foreign key constraint name, named like in the following pseudo-code, can then easily exceed the PostgreSQL maximum constraint name length of 63 characters. When the column names are already minimal, the best recourse that I found was to try to shorten the table names. intercalate "_" $ tableName : columnNames ++ ["fkey"]Perhaps providing a way for users to specify a custom constraint name would be an acceptable workaround? This would allow developers to avoid the issue when renaming tables is not desirable. |
This changes foreign key constraint names to be based on the table and columns of the foreign key, so that they are unique. This is a breaking change.
See #38 for details.