Lack of interest Upgrader suggestion

[Marc]

Please could we have something added to the upgrader that checks at the beginning what columns should be added and whether they have already been added for whatever reason and they shouldnt be there yet.

To explain a bit further. There are 3 posts so far of people who's attach in conversations would not work. This was due to a modification that had added a column that xenforo 1.1 was to add, so because we uninstalled the attachment modification after upgrading it then removed a column that is required by 1.1.

Suggestion:

1. Check version currently being used
2. Check what columns are to be added
3. If any of them columns exist before upgrade, there is a problem so therefore error stating x column on x table has already been added to the database.

Alternative:

1. Check on install of a modification if any SQL statement is inserting a column/table using xf naming convention.
2. On finding, error and state as much

I think this may save a fair amount of headache on upgrade, and also save you guys a fair amount of support that is not really needed.

 

[Mike]

Throwing errors like you want when schema changes are already made will actually cause way more errors - you can see that with the beta 1 xf_identity_service errors.

There's no way to really distinguish between a half completed upgrade and columns that may have been manually added.

 

[Marc]

Throwing errors like you want when schema changes are already made will actually cause way more errors - you can see that with the beta 1 xf_identity_service errors.

There's no way to really distinguish between a half completed upgrade and columns that may have been manually added.

OK, then how about the second one, stopping installers using the same naming convention as xf tables? I appreciate you cant stop people manually adding them, however anyone who manually adds will have at least some experience. Anyone can install a modification with an installer and if those modifications are somewhat stopped from breaking your table structures surely that would be a good thing?