html_url,issue_url,id,node_id,user,user_label,created_at,updated_at,author_association,body,reactions,issue,issue_label,performed_via_github_app
https://github.com/simonw/sqlite-utils/issues/586#issuecomment-1683404978,https://api.github.com/repos/simonw/sqlite-utils/issues/586,1683404978,IC_kwDOCGYnMM5kVriy,9599,simonw,2023-08-18T06:13:46Z,2023-08-18T06:13:46Z,OWNER,"I shipped the view recreating fix in `datasette-edit-schema`, so at least I can start exercising that fix and see if it has any weird issues.","{""total_count"": 0, ""+1"": 0, ""-1"": 0, ""laugh"": 0, ""hooray"": 0, ""confused"": 0, ""heart"": 0, ""rocket"": 0, ""eyes"": 0}",1856075668,.transform() fails to drop column if table is part of a view,
https://github.com/simonw/sqlite-utils/issues/586#issuecomment-1683398866,https://api.github.com/repos/simonw/sqlite-utils/issues/586,1683398866,IC_kwDOCGYnMM5kVqDS,9599,simonw,2023-08-18T06:05:50Z,2023-08-18T06:06:42Z,OWNER,"Options:
- Provide a `recreate_views: bool` parameter to `table.transform()` controlling if views that might reference this table are stashed and dropped and recreated within a transaction as part of the operation. But should that be `True` or `False` by default?
- Read that `PRAGMA` and automatically do that view workaround if it's turned on
- Toggle that `PRAGMA` off for the duration of the `.transform()` operation and on again at the end. Does it only affect the current connection?
- Try the `transform()` in a transaction, detect the `""error in view""`, `""no such table""`error, if spotted then do the VIEW workaround and try again

I'm on the fence as to which of these I like the most. I'm tempted to go with the one which just drops VIEWS and recreates them all the time, because it feels simpler.","{""total_count"": 0, ""+1"": 0, ""-1"": 0, ""laugh"": 0, ""hooray"": 0, ""confused"": 0, ""heart"": 0, ""rocket"": 0, ""eyes"": 0}",1856075668,.transform() fails to drop column if table is part of a view,
https://github.com/simonw/sqlite-utils/issues/586#issuecomment-1683396150,https://api.github.com/repos/simonw/sqlite-utils/issues/586,1683396150,IC_kwDOCGYnMM5kVpY2,9599,simonw,2023-08-18T06:02:18Z,2023-08-18T06:06:31Z,OWNER,"More notes in here:
- https://github.com/simonw/datasette-edit-schema/issues/35#issuecomment-1683392873

Not all Python/SQLite installations exhibit this problem by default!

It turns out this is controlled by the `legacy_alter_table` pragma: https://sqlite.org/pragma.html#pragma_legacy_alter_table

If that PRAGMA is turned on (default in newer SQLites) then `alter table` will error if you try to rename a table that is referenced in a view.

Here's a one-liner to test if it is on or not:

```bash
python -c 'import sqlite3; print(sqlite3.connect("":memory:"").execute(""PRAGMA legacy_alter_table"").fetchall())'
```","{""total_count"": 0, ""+1"": 0, ""-1"": 0, ""laugh"": 0, ""hooray"": 0, ""confused"": 0, ""heart"": 0, ""rocket"": 0, ""eyes"": 0}",1856075668,.transform() fails to drop column if table is part of a view,