Post
Topic
Board Meta
Re: [Guide] Searching effectively
by
tranthidung
on 31/12/2024, 16:18:23 UTC
Please don't delete the bump post above, I love it as a flag of Ukraine.

Some tricks on searching from theymos.
It looks to me like it's working as usual. (Which is to say, not great.)

I use the forum search a lot, but you really need to know how it works and how to deal with its quirks. Some tips:

  • If you search for candy corn, it will return posts with both "candy" and "corn" in their bodies or subjects. The two words are keywords that can be separated, so it would match the phrase "I like candy, but I don't like corn". If you instead search for "candy corn", with the quotes, then the two words have to appear right next to each other as a single phrase.
  • With the default settings, it will only return the last post in a topic that matches, even if more posts in that topic match. To change this, enable the option "Show results as messages".
  • If more than about 5 pages of topics would be returned, then only an arbitrary subset of all results will be returned. To avoid this, setting a message-age restriction is often a good approach.
  • If you use the search box in the upper right while in a topic or board, then it only searches that topic or board (& sub-boards).
  • If one of the words in your search is incredibly common on the forum, like "the", then search will return bad, weird results, or may time-out after hanging for a long time. This applies even if the common word is in a quoted phrase, like "the president". Some decent-looking results may be returned, but posts which should be returned will not be.

Search has a number of quirks, and I'm not going to fix them all right now. In this case, your search didn't work as expected because words appearing in a large percentage of posts are not indexed within posts at all. Currently, there are 181 unindexed words, and "bitcoin" is one of them. So if you search for "bitcoin", it will not even look for posts containing "bitcoin" in their bodies, though it will search topic subjects, and it will highlight "bitcoin" in any posts found via subject-search. As another example, if you search for "bitcoin candy", it will search for messages with [both bitcoin AND candy in their subject] OR [candy in their body], even though it will highlight both "bitcoin" and "candy" in message-bodies in the results. (Furthermore, this will exceed the result limit, so you will actually only get an arbitrary subset of the possible results.)

Next time I reindex posts, probably it would be a good idea to not have any nouns in the list of unindexed words.