Bug: No substring search in backend

[ballaschk]

When searching contacts, documents or images from the main editor, only fully matching words are accepted as search terms in the self-populating results screen. This is true for tags, file names, image descriptions, etc. ...

Example:

• "uh" does not give "uhr.png" as an image result, only "uhr" will
• "stoc" does not give "Stock Image" as an image result", only "Stock" will
• "Mart" does not give Martin B and Martin V as contacts results, only "Martin" will.
• "Sicherheit" will not return documents tagged with "Sicherheitsbelehrung"
• etc.

[donald]

Contacts / Persons seem to work with substring search in the backend.

Unlike Contacts, Documents and Images use the wagtail search for which we currently use the postgres_search backend. The search-function of that backend, however, doesn't seem to support substring searches, so this is an upstream issue. However, the backend is able to index for substring match and the autocomplete-function of the backend can search for partial matches. I could even make that available to Documents and Image (Index and Chooser) by just replacing .search() with .autocomplete(). So we might have dirty a workaround but I'm not yet decided how this hack could be integrated.

[donald]

@ballaschk: Experimental patch is applied on the life server. Please test.

[ballaschk]

Great! It works in the interface that comes up in the rich text editor ("Images", "Documents" buttons).

But not in the "Contact" picker after adding a section with a contact (see below).

[donald]

Thanks for the screenshot. With the above commit, I only fixed "Documents" and "Images" . "Contacts" already work from the page, where you edit the Contacts. So there is the chooser/picker for contacts left. I'll have a look into that.

[donald]

I've made a workaround for the "Contacts" chooser. Test, please. In my test "her" is not matched , I think postgres fulltext index has stopwords for common words like "the" "it" and possible "his" and "her".

[ballaschk]

Worked flawlessly in the instances I needed it. Thanks a lot!