XFES ignores setting searchMinWordLength

Kirby

Well-known member
Affected version
2.2.2
1710077797350.webp

The explain text explanation somewhat suggest that this setting is used with any search system, eg. including XFES.

But this is not the case, XFES ignores this setting which causes all words to be indexed, no matter how short they are.

The description should either indicate that this setting might not be used by a search system other than MySQL full text or other search systems including XFES should respect this setting (Preferred)
 
other search systems including XFES should respect this setting (Preferred)
Is there real value in implementing support for this in XFES? It's not especially difficult but, as Xon points out, this is a major selling point for a lot of people. Adjusting the option appropriately when XFES is installed/upgraded/uninstalled to maintain the current behavior is potentially troublesome too.
 
We've had a case where undesired short word results were returned by ES and our community managers reported it as a bug because it "should not have been returned due to configured minimum word length".

So ... yeah, IMHO this setting really should be respected, at least it should be documented that it might not apply if implementations other than MySQL fulltext are used.
 
Last edited:
I agree that documenting it is important to prevent confusion, but I'm not certain that supporting this could meaningfully improve result quality. I think the only impact would be that shorter terms would be ignored by users who deliberately search for them.
 
As already posted in the initial report:
I am fine with a "documentation fix", but I think it is always better to have choice and I would therefore probably change the implementation so the setting is respected.

You could also add an additional option to XFES config (also needs documentation for the XF setting that it might be ignored):
[ ] Respect Minimum Word Length
If enabled (disabled by default) XFES does respect the XF setting - if not it does not.

=> No issues with install/upgrade/uninstall but the admin still has choice
 
I agree that documenting it is important to prevent confusion, but I'm not certain that supporting this could meaningfully improve result quality. I think the only impact would be that shorter terms would be ignored by users who deliberately search for them.
You get better search results by indexing a "exact"* and "stemming/stop-words" and allowing elasticsearch to search both fields while biasing to the "exact"

*Well near exact, as you still want todo some text normalization.
 
Last edited:
Top Bottom