v0.3.0 - Mupdf Rewrite

0.3.0

• Update ratatui(-image) dependencies
• Enable Ctrl+Z/Suspend functionality
• Rewrite with mupdf as the backend for much better performance and rendering quality
• Support easy inversion of colors via i keypress
• Support for filling all available space with f keypress
• Change help text at bottom into full help page

With regard to the performance, these are the kind of changes we're seeing:

   Compiling tdf-viewer v0.2.0 (/home/june/Documents/coding/personal/tdf)
    Finished `bench` profile [optimized] target(s) in 1.70s
     Running benches/rendering.rs (target/release/deps/rendering-63f7619046d0b14f)
render_full/benches/adobe_example.pdf
                        time:   [43.921 ms 46.470 ms 49.137 ms]
                        change: [-9.8772% -2.8599% +4.2524%] (p = 0.44 > 0.05)
                        No change in performance detected.
Found 2 outliers among 40 measurements (5.00%)
  2 (5.00%) high mild

render_full/benches/example_dictionary.pdf
                        time:   [928.86 ms 933.67 ms 938.67 ms]
                        change: [-41.719% -41.373% -41.049%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 1 outliers among 40 measurements (2.50%)
  1 (2.50%) high mild

render_full/benches/geotopo.pdf
                        time:   [399.14 ms 403.13 ms 407.19 ms]
                        change: [-64.450% -64.040% -63.656%] (p = 0.00 < 0.05)
                        Performance has improved.

render_first_page/benches/adobe_example.pdf
                        time:   [20.232 ms 21.225 ms 22.265 ms]
                        change: [-24.649% -19.961% -14.677%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 40 measurements (7.50%)
  1 (2.50%) low mild
  2 (5.00%) high mild

render_first_page/benches/example_dictionary.pdf
                        time:   [15.449 ms 16.222 ms 17.033 ms]
                        change: [-8.1274% -1.7836% +5.7176%] (p = 0.61 > 0.05)
                        No change in performance detected.
Found 1 outliers among 40 measurements (2.50%)
  1 (2.50%) high mild

render_first_page/benches/geotopo.pdf
                        time:   [47.447 ms 50.062 ms 52.744 ms]
                        change: [+61.358% +73.485% +86.469%] (p = 0.00 < 0.05)
                        Performance has regressed.
Found 8 outliers among 40 measurements (20.00%)
  3 (7.50%) low mild
  4 (10.00%) high mild
  1 (2.50%) high severe

only_converting/benches/adobe_example.pdf
                        time:   [3.0548 ms 3.4825 ms 3.9492 ms]
                        change: [-56.335% -46.955% -36.602%] (p = 0.00 < 0.05)
                        Performance has improved.

only_converting/benches/example_dictionary.pdf
                        time:   [223.93 ms 231.58 ms 239.54 ms]
                        change: [-62.939% -61.546% -60.341%] (p = 0.00 < 0.05)
                        Performance has improved.

only_converting/benches/geotopo.pdf
                        time:   [65.724 ms 70.457 ms 75.330 ms]
                        change: [-42.242% -37.122% -32.047%] (p = 0.00 < 0.05)
                        Performance has improved.

Obviously, this all looks quite good, with the exception of render_first_page/benches/geotopo.pdf. This tests how long it takes for the first page of the geotopo test file to appear to the reader, and it's the only test that had any sort of regression (and it's a quite sizeable regression, too).

This is obviously something I'll be looking into in the future, but I don't think it's really any cause for concern right now. Looking at this test, 20% of its recorded measurements were outliers. This is a huge red flag by itself, and when we also look at the other benchmarks that use this file, we see that it's even more of an outlier (since all other benchmarks involving this file improved quite dramatically).

I think this test is just very noisy and unreliable, on top of perhaps causing some sort of edge-case in the mupdf rendering code. It'll be looked into, but I don't think many (if any) users of tdf should see this sort of performance regression. If you do, please file a bug :)