Re: custom-set-variables considered harmful

[Elias Mårtenson]

On 28 November 2017 at 22:12, Philippe Vaucher <address@hidden> wrote:

Note that as long as the `setq` is within the magical

custom-autogenerated-user-settings form, it will behave correctly even
if the variable has a setter function.  The difference will only affect
those users who take this code and then copy it elsewhere.

But setq doesn't have any concrete benefit other than familiarity to people who don't understand custom. I've spent more time than I wish on Freenode #emacs teaching people about custom. If c-s-v was used consistently that effort would be made a lot easier. 

The problem I have with c-s-v is that I don't see its benefits for configs like this:

 

[ lines of ‘setq’ invocations on ido variables ]

Can you maybe enlighten me about what problem c-s-v solves that would have me start to use it?

Your explicit ‘setq’ calls are perfectly fine as long as the variables you set do not have a setting function

assigned to them. However, if they do have setter functions, then the behaviour will not be correct.

It has been suggested that you manually look at the ‘defcustom’ definitions of the variables so that

you know which ones doesn't work with ‘setq’.

The only reason people can get away with not doing this is because there are so few custom variables

that use setter functions.

Thus, if you use ‘setq’ instead of ‘custom-set-variable’ you need to manually check every single

variable to ensure they don't have setter functions assigned. If you want to be thorough (nobody is)

you also need to check them every time your modules are updated. This is not just a theoretical

situation. It happened in gnu-apl-mode, a library for which I am the maintainer. In it, the

variable ‘gnu-apl-mode-map-prefix’ got a setter function after the fact.

Or, you could simply use ‘custom-set-variable’ everywhere and spare yourself the trouble.

Regards,

Elias