I did a bunch of testing of various ansi escape codes (including
italics) with all three char atlas implementations (none, static,
dynamic), with and without `drawBoldTextInBrightColors`, and with and
without `enableBold` to test all of the edge cases that I could.
Adjacent cells on the same row sharing the same background color will be
drawn using the same `fillCells` call. This should make drawing an
application with a solid or mostly solid background color (e.g. vim)
faster.
I set vim to draw the background color for a file, and started scrolling
through it. Before, _drawBackground was taking around 6ms per frame.
After this commit, it was taking around 0.5ms per frame.
Based on prior experience, I'd expect these results to be more drastic
for larger terminal windows. The demo's window is pretty small.
I'm primarily interested in doing this because it'll allow me to
optimize the background rendering in later diffs, but this may also make
it possible to fix some minor rendering issues.
Right now, if we draw a single-width character that overflows its cell's
bounds, and the character to the right of it has a background, that
background may cover up the first character's foreground. By drawing the
background in a separate pass, we can avoid those cases.
There's still some issues with how dirty regions are computed that makes
rendering stuff like that flaky, but this at least gets us closer to
"correct" rendering.
Other terminal emulators (e.g. alacritty) render the foreground and
background in separate passes:
https://github.com/jwilm/alacritty/blob/1b7ffea/src/renderer/mod.rs#L766
This limits the amount of damage a pathological program can cause in
thrashing the LRU cache.
I ran the script described here:
https://github.com/xtermjs/xterm.js/pull/1327#issuecomment-374003960
Without this patch, the entire output would get drawn in one frame that
took about 140 ms to draw. With this patch, it takes about 60 ms. It's
still nowhere as good as the static or none atlas implementations, but
it's not terrible, like it was.
This should have minimal impact on real-world applications: it'll just
take a little longer for the atlas to warm up.
The implementation is a little hacky, since it involves throwing some
code into TextRenderLayer that violates the separation of concerns, but
we can clean this up later.
This mostly worked before, but clearColor made the assumption that the
colors were always in #RRBBGG format. See the discussion here:
https://github.com/xtermjs/xterm.js/pull/1327#issuecomment-374812842
This changes the internal representation of a color so that we hold onto
into an object containing the original css string along with an RGBA
value encoded as a 32-bit uint. clearColor can then use that RGBA
representation.
Additionally, this deduplicates some code between ColorManager and
Terminal. Terminal was generating the 256 ansi colors the same way
ColorManager was, so this just makes Terminal use the same constant.
Two parts were commented out:
- _writeStopped: This is because _xoffSentToCatchUp references this variable
and the feature is sort of half implemented/disabled.
- _clearChar: This is because some comments still reference this, it's part
of the work needed to do better line redrawing.
- Ensures that everything is only defined in one place, and everything
imports from that place.
- Moves CHAR_ATLAS_CELL_SPACING into a subdirectory of shared/ so that
CharAtlasGenerator can pull from it.
This addresses the comments on
https://github.com/xtermjs/xterm.js/pull/1307/files/454fd4bfb1ab5c9ece
The only thing `didCharSizeChange` was used for was to figure out if we
need to refresh the character atlas or not. However, `acquireCharAtlas`
already inspects the atlas owned by this terminal, and if it matches, it
avoids generating a new atlas.
So if `didCharSizeChange` is true and it's not needed, it quickly bails
out as a no-op. But if `didCharSizeChange` is false, and it was needed
(possible given the complexity of testing all of these edge cases), it
could introduce a bug.
I think it makes sense to just get rid of `didCharSizeChange`, and just
always call `acquireCharAtlas`.
* Measure characters and thread them if they are overlapping
* Fix whitespace
* Use charData and ctx.save/restore, code cleanup
* Fix trailing whitespace
* Corretly inverse foreground and background. Be more conservative when clearing
* add missing space
* Re-enable optimisation with better checks
* Fix insversed check
Jörg Breitbart
Daniel Imms
Jeff Principe
Benjamin Woodruff
Linus Unnebäck
Daniel Imms
Brandon Bayer
CHaBou
Bruno Ribeiro
Thomas Zilz