Is it expected to rank match at middle or end of word higher than start of word?

[jn64]

trafficstars

Version: 20240414.842 (melpa) Commit: 864d07cadb0ecbbf6c296db6ecfbb38358436e21 No dynamic module

(require 'hotfuzz)
(setopt completion-styles '(hotfuzz)
        completion-category-defaults nil
        completion-category-overrides nil)

Test 1, M-x fop TAB (uppercase indicates matched characters):

FOrward-sexP
FOrward-Page
...

Test 2, M-x dpk TAB:

Diff-sPlit-hunK
Diff-aPply-hunK
Describe-PacKage
Describe-Personal-Keybindings
...

This was surprising based on the description "ranks completions that match at word; path component; or camelCase boundaries higher."

At least in test 1, end of word is a boundary. But in test 2, I expected describe-personal-keybindings first.

[axelf4]

This is not ideal, but it is due to there not being a gap at the end if the last characters of a completion are matched, and gaps being penalized the most. E.g.

FOrward-sexP
      1     gap
FOrward-Page
     1    2 gaps