UX for adding codelists to a project is poor

[Jongmassey]

To add a codelist from OpenCodelists to a study a user currently has to:

1. Identify the codelist required on OpenCodelists
2. Copy the part of the URL after https://www.opencodelists.org/codelist/ being sure not to include the trailing /
3. Paste into codelists/codelists.txt
4. Run opensafely codelists update
5. Note the path of the codelist csv that has been downloaded by this command
6. Inspect the csv/the codelist on OpenCodelist to note the key field name(s)
7. Use the codelist_from_csv() function in their ehrQL with the field name(s) and csv file path

That's rather a lot of steps, and opportunities for things to go wrong.

Improvement of 5-7 could require changes to ehrQL or perhaps generation of code snippets, but I feel 2-4 are ripe for some simplification by changing the behaviour of the CLI.

[Jongmassey]

For a while I've thought we could provide an easier means of adding codelists, perhaps via an add command, and @lucyb appears to have reached a similar thought so I propose this as a potential implementation of this.

In my proposed implementation the user:

• doesn't need to modify the opencodelists URL
• doesn't need to modify the codelists.txt file
• only needs to call one CLI command

The last of those may require some further thought as there are sometimes-unexpected implications/side-effects of calling opensafely codelists update e.g. updating previously-downloaded codelists, and removing and local codelists which may have (erroneously) been placed in the codelists/ folder. However, I cannot imagine a circumstance in which a user would add a codelist to codelists.txt and not then run opensafely codelists update.

[Jongmassey]

Any such changes to the the CLI would require updates to the documentation and to (e.g.) copiloting guidance.