quitesimpleorg / looqs Goto Github PK

Consider adding completion in main window lineedit:

• History
• Filter names
• Automatic paranthesis

Currently we won't add the same file twice simply by matching the path, but if there is a copy somewhere else, we will add that one too and also reindex it.

Thus, either:

• Don't store duplicates
• At least reuse the index.

and:
Maybe an option to prune duplicates afterwards.

While generally the previews are sorted by relevance, the order they are shown in the previews tabs also depends on which thread is fastest to generate a preview. Practically we therefore do not always match sqlite's ordering.

So we should do some sorting of the previews, however we cannot wait till all are finished either.

They can be improved for large files. We may need to create a snippet that contains all words one way or another

Consider:

• in search results, right click on document, then "show previews only for this document" or something
  

• in previews, right click on a preview, then "show all previews for this document...". Alternatively, a combobox with all filenames     within the current previewables. Or a list.
  

• in previews: Options to not show those where no words have been marked
  

Currently, text is text. There is no context information.

It would be great if we could to something like: "c:(word) type:heading" to search in headings of documents.

For PDFs, we may simply extract the index to do that, at least, as a first step.

We should either convert to contentless FTS approach or use the content to speed up plaintext previews. Currently it's wasted space.

Distributions make it available slowly it seems, so it must be on our radar now

It's slowly becoming quite the god class...

Perform OCR on images / screenshots, store in index etc.

Optionally emit every occurance of search terms in plain text files, i. e. no truncation.

Currently, search results that are not previews are just a list. It might be better to offer a tree-view, so it's easy to see which files are in which directory.

I. e. like QDirStat, but should expand all by default.

Things to think about:

• Implement custom sqlite tokenizer?
• Or do it before passing the content to sqlite?
• What about prhase queries?

Currently, allows read on all /.

We theoretically know the render targets, we could launch a subprocess restricting the files we are allowed to read to those (+ a few other such as libs).

Automatic tests, particularly for queries