Inconsistent Search Results Behaviour

[zachmandeville]

Title:
On the envoy docs, search results are inconsistent across similar search terms.

Description:
I noticed when searching for ip tagging, that I could only find the page when i matched text in the qualified names referenced on the page. Other pages I could find by matching any text on the page.

Repro steps:
.
I reproduced the below behaviour on firefox, chrome, and safari. I did not test on all versions, but reproduced using latest, 1.17.1, 1.16.1, and 1.13.1

• in the search field of the docs enter any of these terms: "ip tagging", "ip tag", "ip" and note the results.
• enter "iptag" and note the results.
• enter "kafka broker", "kafka", "kafka b", "emitting results", then note results.

The first set of terms will not bring up the ip tagging page. "iptag" will return the page by matching on the qualified names in the page. Any of the last terms will bring up the correct kafka broker page among its results.

Admin and Stats Output:
n/a

Config:
n/a

Logs:
n/a

Call Stack:
n/a

[zachmandeville]

cc: @phlax

[phlax]

@zachmandeville thanks for reporting

ive noticed that the search is not very good on a number of occasions - tbh ive never been that impressed with sphinx's search

there are a few search settings set in the conf.py i guess its worth playing around with these to see if any help to improve results

i think there is also a search adapter that you can add to further customize, which may be useful in resolving specific issues - https://www.sphinx-doc.org/en/master/usage/advanced/websupport/searchadapters.html

if you were to experiment with these and PR any improvements i would be very up for reviewing - otherwise i will take a look at this as i have time

[zachmandeville]

Thank you for these resources! I'll do some experimenting as time permits.

[daixiang0]

Upstream fix as sphinx-doc/sphinx#8070
PR as #16186

[phlax]

@daixiang0 it would be good to know whether this bug is now fixed by upgrade to sphinx 4

[daixiang0]

I update the pr and I test in latest doc website with "ip tag", it works.

[phlax]

great, thanks @daixiang0