Add a search engine into Plume

[trinity-1686a]

Add support for search through tantivy
fix #149

This is mostly done, current issues are:

• query parser might be a bit intolerant, we should probably implement our own
• closing Plume leave the database blocked, currently you have to run plm search unlock -p to unlock it.
• frontend is done by me (which means it's probably not beautiful), and is not translated
• advanced search is working, but yet undocumented (i.e. filtering by date could be done if you know enough tantivy, but I don't expect common user to)
• As it has grown more than I anticipated, I should probably add some more tests

After upgrading to this, you will have to run plm search init -p <path to plume wokingdir> to initialize search database, otherwise Plume will panic at launch

[codecov]

Codecov Report

Merging #324 into master will increase coverage by 5.55%.
The diff coverage is 65.36%.

@@            Coverage Diff             @@
##           master     #324      +/-   ##
==========================================
+ Coverage   22.42%   27.97%   +5.55%     
==========================================
  Files          55       61       +6     
  Lines        4883     6419    +1536     
==========================================
+ Hits         1095     1796     +701     
- Misses       3788     4623     +835

Impacted Files	Coverage Δ
src/routes/user.rs	0% <0%> (ø)	⬆️
src/routes/mod.rs	0% <0%> (ø)	⬆️
src/routes/posts.rs	0% <0%> (ø)	⬆️
plume-cli/src/search.rs	0% <0%> (ø)
src/api/posts.rs	0% <0%> (ø)	⬆️
src/main.rs	0% <0%> (ø)	⬆️
src/inbox.rs	0% <0%> (ø)	⬆️
src/routes/likes.rs	0% <0%> (ø)	⬆️
src/routes/comments.rs	0% <0%> (ø)	⬆️
src/routes/reshares.rs	0% <0%> (ø)	⬆️
... and 35 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 74c398d...6d83d5a. Read the comment docs.

[trinity-1686a]

Small todo list of things that are not strictly needed atm, but should still be done:

• split plume-models/src/search.rs into multiple files, under a new directory : it's "only" 670 lines, but they are very dense, and although all is related to search, having a sort of database connection, a query parser, and a tokenizer in the same file seems too much
• comment the code, especially the parser
• maybe make some part less generics but more logical. Currently if searching for author user1@plume1.org or user2@plume2.org, a post from user2@plume1.org would match.
• add some tests at least for the parser
• use ructe for the template, and maybe change how js change form on submit (currently going backward in history and re-submit has some issues)

[fulmicoton]

query parser might be a bit intolerant, we should probably implement our own

@fdb-hiroshima We have a dangling PR to add a lenient mode to search.

i.e. filtering by date could be done if you know enough tantivy, but I don't expect common user to)

@fdb-hiroshima Let me know if you have any question about tantivy.

[trinity-1686a]

@fulmicoton Both "issues" are fixed (you can see they are crossed out), I've implemented a parser which fit with our exact needs (we know more about what each field represent than Tantivy can).

Concerning dates, I don't have any question, it's just how we give date to tantivy is not evident to the end user, but tantivy can't change that. We store date as the result of num_days_from_ce(), basically the number of days since some point. This is because Tantivy's RangeQuery iterate over the terms within the range, so I tried to have the smallest ranges possible, with as few useless/empty elements as possible (iterating over 30 days vs 2,592,000 seconds, one seems faster than the other).

I really like your crate, it's very complete yet relatively easy to use. The only issue I had with it was about BooleanQuery, I had a hard time understanding how to build theme. I was looking for some kind of function, and took a lot of time to figure out it implement From. Maybe it could just simply be mentioned in the documentation of BooleanQuery::new_multiterms_query, "to create a BooleanQuery from Box<Query> use From<Vec<(Occur, Box)>> instead" or something like that

[fulmicoton]

Opened tantivy-search/tantivy#446 to make creation of BooleanQuery easier.

Your insight about range is correct. Eventually tantivy will be a bit smarter than what it does right now, but too be honest this is pretty low priority.

Looking forward to seeing Plume grow.

[elegaanz]

Seems to work well! Thank you! (and thank you @fulmicoton for tantivy as well)

[trinity-1686a]

@BaptisteGelez in what order do we merge search ructe and the dependency upgrade?

[elegaanz]

@fdb-hiroshima I think we could merge the search first, then rebase ructe on master and convert the search page to Ructe, and finally the dependencies? I don't know if there is a better way to avoid conflicts, but it seems to be the more logical way to do it for me…

[trinity-1686a]

I'll do so

[trinity-1686a]

There is one thing I forgot. There is nowhere a link to /search
And there is also no instruction to run plm before

[elegaanz]

You're right, but I'll a a link the header when converting the template to ructe. And updating the docs can be done in another PR.

[trinity-1686a]

if you want you can make a small form with an <input name="q">, it is enough to make a simple query, and would probably fit better with the ergonomic people expect (being able to do a basic search from anywhere, and an advanced search if the result were inconclusive)