Feedback

[wasamasa]

multi-benchmarks.el

• mu- is a pretty bad namespace prefix due to confusion with mu4e (where the native parts are known as mu).
• Use cl-lib consistently, cl-lib-hightlight helps.

multi-methods.el

• Please don't do lexical binding hacks, instead use it in all files you write. A template for them can help. Tell people using your library likewise.
• concurrent access can happen in process filters, irc.el has an instructive comment: https://github.com/jorgenschaefer/circe/blob/6ccd4b494cbae9d28091217654f052eaea321007/irc.el#L122-L133
• Comment headers of the ;;* foo --- *;; variety are anything but standard. Use ;;; foo instead and check out the coding guidelines in the reference.

multi-patterns.el

• Yes, interactive is special, it can accept code that will be evaluated at a later time. No idea about declare.
• Consider using buttercup if ERT causes you massive pains.
• Mutating list literals is a bad idea in any case, don't do it: https://web.archive.org/web/20170106022314/http://shift-reset.com/blog/2014/2/28/List%20Mutation%20Public%20Service%20Announcement/
• ? doesn't translate to 32, it's char syntax for a space (which is better expressed as ?\s anyway). You'd be better off escaping the question mark instead, that would make it a symbol.
• Generally speaking, consider not translating mu-patterns to pcase and roll your own thing instead. This would be a greater undertaking, but teach you a lot more and most certainly avoids degenerate pcase cases. There are a bunch of implementations in Emacs Lisp, Scheme, Clojure, the better ones reference papers.

multi-prelude.el

• This file will most certainly not fly because its unprefixed stuff goes way beyond shimming. MELPA's packaging standards are pretty strict here.

[vkz]

thank you @wasamasa

Since people have been sending technical feedback by email, here on github and on Reddit, I'll try to compile them all here lest I forget what I need to do as I get to, well, doing it.

From email communication with Stefan M:

• mu-seq-pattern-force-list can't meaningfully be a "defcustom" because
  changing the setting will break existing code that relies on
  the other setting of that var, and making the code agnostic to the
  value of this setting means that changing this setting will have no
  visible effect for the end-user.

This one needs documentation change:

• mu-function? is not "like functionp" since it doesn't operate on
  values but on expressions.

Reddit:

Similar to what @wasamasa reports: namespacing, cl-lib, etc