Improve correctness and function declarations

[swapneils]

Correctness changes:

• join now treats a nil separator as the empty string, rather than the string "nil", conforming to the semantics in the README of not converting nil to "nil".
• containsp now automatically returns nil if either string input is nil.
• alphanump and similar now use the correct regex checks for multi-line strings, rather than accepting an input if it has any line matching their specification.

Optimization changes:

• Multiple non-recursive functions have been inlined
• Multiple functions have ftype declarations
• In local testing, there is a ~0.1s reduction from ~1.05s to ~0.95s in (time (dotimes (i 100) (with-output-to-string (throwaway) (let ((*standard-output* throwaway)) (asdf:test-system :str))))), despite the new tests in this version.
  • Note: Most of the tests use constantp inputs, so performance improvements here derive mainly from the inlining; they do not represent benefits from the ftype declarations in user-code.
  • Note: The comparison was specifically with the contents of https://github.com/vindarel/cl-str/pull/128, but that still shows an improvement over the original.

[vindarel]

excellent, thanks, it looks good, though I'll take more time to test in the wild, because merging bad type declarations can be very annoying (and imagine if Quicklisp releases in-between our quick fixes). I encourage anyone to try this PR.

[vindarel]

hey btw I have compilation warnings "deleting unreachable code" for the pad-* functions:

; in: DEFUN PAD-LEFT
;     (STR:PAD STR::LEN STR::S :PAD-SIDE :LEFT :PAD-CHAR STR::PAD-CHAR)
; --> BLOCK LET IF FLET STR::%PAD-CENTER BLOCK MULTIPLE-VALUE-BIND 
; --> MULTIPLE-VALUE-CALL FLOOR - 
; ==>
;   STR::LEN
; 
; note: deleting unreachable code

; --> BLOCK LET IF FLET STR::%PAD-RIGHT BLOCK CONCATENATE 
; ==>
;   STR::S
; 
; note: deleting unreachable code

[…]

Isn't there an issue when the arguments len s pad-char are the same than the mother pad function?

[vindarel]

I would do this:

(declaim (inline pad-left))
(defun pad-left (%len %s &key (%pad-char *pad-char*)) ;; %len
  (pad %len %s :pad-side :left :pad-char %pad-char))

;; same for pad-right

or remove their inline declaration. @swapneils ?

[swapneils]

Thanks for looking into the compile failures! I'll aim to check that change this weekend and see if a (few-months-old) source build of SBCL raises anything else. Been pretty busy with work and work-adjacent stuff the last few months.

[vindarel]

No more pad-* compilation warnings with a newer compiler (SBCL 2.5.8).