890323 - Define and Implement search syntax for /search

Status: RESOLVED FIXED

(Webmaker Graveyard :: webmaker.org, defect)

Description

[Cassie McDaniel [:cassiemc]]

Eventually we'll want to search by multiple words, so something like "teaching kit". Right now it looks like it is searching "teaching" and "kit" – it even automatically inserts a comma between the two when it tells me there are no results.

We want each search term, or tag, to be separated by comma – but not automatically, the user should have control over this. We can follow google's model, allowing terms to be searched together by enclosing them in quotes.

Comment 1

[Cassie McDaniel [:cassiemc]]

From Chris:

> (In reply to Cassie McDaniel [:cassiemc] from comment #1)
> > Cade, does the Make API support multi-word tagging at the moment? I believe
> > I filed this as a bug last week. We might want to tag something "15-30
> > minutes" for example.
> 
> I'm not sure if it will. I'll investigate.

Chris, here are some example tags that are desired by the mentor team: https://etherpad.mozilla.org/Teaching-kit-tags

Comment 2

[Chris DeCairos (:cade)]

(In reply to Cassie McDaniel [:cassiemc] from comment #1)
> 
> Chris, here are some example tags that are desired by the mentor team:
> https://etherpad.mozilla.org/Teaching-kit-tags

I've confirmed that all the tags in that etherpad are working. any tags that contain commas will be problematic.

Comment 3

[Chris DeCairos (:cade)]

Is this bug looking for more powerful search syntax on Webmaker.org/search?

Comment 4

[Cassie McDaniel [:cassiemc]]

So even tags with ampersands – "Sharing & Collaborating" and strange punctuation like "<30 min", "30 min to 1.5 hr" (or alternately, "30min-1.5hr"), and "1.5 hr+" – those all area fine?

What is the current search syntax, does it mirror google's, so separating each tag by comma or do they have to be grouped by quotation marks?

> Is this bug looking for more powerful search syntax on Webmaker.org/search?
I'd say Yes, that's it exactly. We want to be able to bring back valid results for users who might have a variety of ways of searching for awkward-looking tags.

Comment 5

[Chris DeCairos (:cade)]

punctuation, excluding ',' and ':' should be fine. Commas cause problems because the make editor assumes they're tag delimiters.

Currently on the website, it searches for what ever iss in the search input box as a whole tag. It's not tokenizing it and applying each word as a tag to filter for.

ie. [hello world] in the search input will pass the contents as one tag, rather than searching for "hello" OR "world"

Clearly defining and documenting our search syntax (and making it easily available to users) will be important.

I've also noticed a weakness in the search API. It's using "filters" when querying Elastic Search. This poses problems because I think it's taking search terms too literally.

Comment 6

[Cassie McDaniel [:cassiemc]]

Chris, can you edit this bug title to whatever it is that needs to happen to make the Search *better*? Better to me means it should find better, more relevant results. If fixing the API filter weakness problem will help, great!

We should file a separate bug on documentation, as I agree that should be upfront.

And just to make sure I understand, [hello world] literally searches the phrase "hello world" – and if I put [Hello, world] then it searches for both "hello" and "world" - is that right?

Thanks for the explanation!

Comment 7

[Cassie McDaniel [:cassiemc]]

Cade and I discussed the expected behavior for this.

[open source] should search the phrase "open source", NOT the separate words "open" and "source". Yes, this means [open source kittens] would search "open source kittens". In order to search both "open source" and "kittens" separately, users would have to write: [open source, kittens], so each search term is separated by a comma.

Moved the documentation task to bug 893025.

Comment 8

[Chris DeCairos (:cade)]

Attachment: https://github.com/mozilla/webmaker.org/pull/267

Splits tag searches only on commas. Trims the split tags, and fixed a crash bug where if the tag began with "#" the app would die.

Comment 9

[Dale Karp (:daleee)]

Comment on attachment 775850 [details] [review]
https://github.com/mozilla/webmaker.org/pull/267

Tested locally, works as expected.
Code r+

Comment 10

[[github robot]]

Commit pushed to master at https://github.com/mozilla/webmaker.org

https://github.com/mozilla/webmaker.org/commit/bf6ab9c72ff1f257f57e5617c917776f9a53e30f
Fix Bug 890323 - Tag search on /search separates on commas only. crash bug fixed for tag searches containing '#' at the front of the string