Feature: Ability to Drop columns

Use case: Removing columns that aren’t needed any more or to allow removal of columns that were accidentally added.

New columns could be added very easily, typos for example:

UPDATE mytable SET someMispelledColumn = ?

Feature description: Allow us to drop columns using the ALTER command, so for example:

ALTER TABLE mytable DROP COLUMN mycolumn;

[celaus]

Hi @bputt - thanks! We already have that in our backlog, I’ll add your upvote to it 👍

[mfussenegger]

From a developer’s perspective, both array _ count(#882) and drop column(#4457) are the most important and relatively easy to implement features

If you think those are easy to implement, we’re happy to accept contributions and can also give you some pointers how to do that if you’re interested.

A slow version of array_count would indeed be rather simple. All one had to do is to implement a new scalar function. Taking a look at one of the PRs that added other scalar functions should be a good starting point. The reason we didn’t add it already, other than lack of time, is that a common use case for it is to query for all records where “an array is not empty” or “an array has more than 1 element”. Implementing array_count as a simple scalar function would lead in a slow query execution and we’d need to introduce an option for an optimized query generation for this common use-case. Otherwise we wouldn’t meet the expectations most people likely have. This makes this “not easy”.

Drop column is even worse once you ask yourself questions like “what to do with the already indexed data” and “what to do if a column with the same name but different type is added later on”. Possible answers to these questions are not easy.

On a more general note. We see that this is often requested and agree that it’s useful and important functionality. We just didn’t get around to do it properly yet.