With HISTORY_SUBSTRING_SEARCH_ENSURE_UNIQUE, it can take a while to compute
all unique search results, which would normally keep the shell unresponsive
during this computation.
This commit separates the retrieval of all history matches from the
computation of the matches that need to be presented to the user (unique
matches only if HISTORY_SUBSTRING_SEARCH_ENSURE_UNIQUE is set). The
computation of the matches needing to be displayed is then done lazily,
and only when the user requests a new previously undisplayed result.
This keeps the response time fast while still enabling us to filter for
unique search results.
If HISTORY_SUBSTRING_SEARCH_ENSURE_UNIQUE is set to a non-empty string,
then all displayed search results will be unique even if the shell
option HIST_IGNORE_ALL_DUPS is not set.
We used to do a history search when the query string was empty, which
resulted in all history indices being returned, but they are never
actually used as _history-substring-search-up-history and
_history-substring-search-down-history will just act like ZSH up-history
and down-history if the query is empty. Removing this unnecessary lookup
should speed up the empty query case, especially if you have a long
history.
This is a pure refactoring patch and does note change any functionality.
We remove an unnecessary reversal of the history match array by swapping
the enumeration order of match indices. Older entries now have higher
internal indices than younger entries, and hence we start from 1.
Also removed one unreachable elif clause each from search-up and
search-down.
Geza Lore