search

shreevatsa / chaya

0 stars 0 forks source link

PM save/load #4

Closed shreevatsa closed 3 months ago

shreevatsa commented 3 months ago

Before getting further into the code, would be good to implement the feature to:

Part of this is UI work, and part is the actual serialization / deserialization.

shreevatsa commented 3 months ago

Let's do a state machine.

Start state = nothing uploaded.

stateDiagram-v2
    [*] --> pdf_picked
    pdf_picked --> pdfjs_initialized
    pdfjs_initialized --> pdfjs_rendered_1
    pdfjs_rendered_1 --> pdfjs_rendered_2
    pdfjs_rendered_2 --> pdfjs_rendered_3
    pdfjs_rendered_3 --> [*]
shreevatsa commented 3 months ago

Next comes the .sc file:

stateDiagram-v2
    [*] --> pdf_picked
    pdf_picked --> pdfjs_initialized
    pdfjs_initialized --> pdfjs_rendered_1
    pdfjs_rendered_1 --> pdfjs_rendered_2
    pdfjs_rendered_2 --> pdfjs_rendered_3
    pdfjs_rendered_3 --> [*]
    pdf_picked --> sc_picked
    pdf_picked --> sc_new
shreevatsa commented 3 months ago

And the PM state:

stateDiagram-v2
    [*] --> pdf_picked
    pdf_picked --> pdfjs_initialized
    pdfjs_initialized --> pdfjs_rendered_1
    pdfjs_rendered_1 --> pdfjs_rendered_2
    pdfjs_rendered_1 --> ocr_1
    pdfjs_rendered_2 --> pdfjs_rendered_3
    pdfjs_rendered_2 --> ocr_2
    pdfjs_rendered_3 --> pdfjs_done
    pdfjs_rendered_3 --> ocr_3
    pdfjs_done --> pm_done
    pdf_picked --> sc_new
    sc_new --> ocr_1
    ocr_1 --> ocr_2
    ocr_2 --> ocr_3
    ocr_3 --> ocr_done
    ocr_done --> pm_done
    pdf_picked --> sc_picked
    sc_picked --> pm_done
    pm_done --> [*]
shreevatsa commented 3 months ago

I'm not sure the above is the only way: we may instead choose to start OCR only after all pages are rendered.

It's more that there are a few separate tracks:

shreevatsa commented 3 months ago

I think it makes sense to OCR the canvas element directly, as this is supported: https://github.com/naptha/tesseract.js/blob/master/docs/image-format.md

shreevatsa commented 3 months ago

Made some changes and page is freezing.

See

shreevatsa commented 3 months ago

The freezing was fixed (with 0f8eaa05669870c4c626c24b6477d206bb1b98ef) and the saving is probably fine; what's left is to load.

shreevatsa commented 3 months ago

What do we load? Currently, the save (not very well thought-out) is view.state.doc.toJSON(). So we'd need to be able to create the PM from this doc?

shreevatsa commented 3 months ago

I think I'll say that save/load are working, and close this. Can create new issues for OCR, and proceed from there.