Fixing and regulating certain types of search fields across KF5 apps

[Eike Hein]

Hi,

we currently have numerous dynamic search fields across KDE
apps that, for performance reasons, disregard input shorter
than a particular QString::length(). Examples are KMail and
some things in KRunner. It's a common pattern.

This is very English-centric behavior; many scripts are
more complex than this, in particular CJK ones. In Korean,
a single character may be composed of multiple letters and
form a complex syllable. Most Korean given names are two
characters, many useful words are only one character.
The situation in Japanese is similar - it's written using
a mix of syllabaries and ideograms. The check for 3 has
good intentions but breaks real use cases for these users,
such as finding mail by name.

Addressing this algorithmically is not all that hard. When
normalizing to Unicode NFD, those complex Korean characters
turn into a sequence of combining characters for the indi-
vidual letters. Syllabic and ideographic code points are
assigned the character class Lo in Unicode, which is info
available via QChar. An NFD pass + counting Lo chars as 2
would be simple and effective enough for this purpose.

The way I'd like to see this fixed is placing an API like
isMinimumSearchableLength(QString) somewhere in KF5 and to
start using it everywhere. Realistically developers won't
find this on their own, but it means I could go around and
patch up all these places without copying code around.

Which framework would be a good place for this?


Cheers,
Eike