The Font class now remembers the results of kerning lookups in a
HashMap. This fixes an issue where text-heavy UI (like WidgetGallery)
would lag when using a UI font with kerning data.
Instead of testing all possible code to find the good symbol, we use a
lookup table to directly find the expected symbol. This method is used
by most Huffman decoder (gzip or libjpeg-turbo).
In order to use the correct key when peeking a constant number of bits
from the stream, we generate duplicates in the table. As an example, for
the code 110, all entries with that pattern 110***** will be set to
110's symbol. So, when you read this code plus garbage from following
codes, you still find the correct symbol.
Advantages of encapsulation are really obvious here:
- Put related code together
- Prevent external functions to modify the object
- Abstract the implementation
No functional changes intended.
A compressed ALPH chunk is a lossless webp bitstream, but without
the 5 byte "header" that stores width, height, is-alpha-channel-used
(it never is for an ALPH chunk since the ALPH chunk gets the alpha
data out of the lossless webp's green channel), and version fields.
For that reason, this cuts decode_webp_chunk_VP8L() into the
header-reading part and the remaining part, so that the remaining
part can be called by the ALPH reading routine.
Lossy webp files with a (losslessly) compressed alpha channel can
be found in the wild. Since we can't decode lossy webp data yet,
change the `#if 0` in decode_webp_chunk_VP8() to `#if 1` to test this.
ALPH chunks are only used to give lossy webp frames an alpha channel,
and lossy decompression isn't implemented yet. So this can currently
never be hit in practice -- but for debugging and testing, I put in
some code behind `#if 0` for now that fake-decompresses a lossy webp
frame by returning an empty bitmap.
But this also doesn't implement compressed ALPH chunks yet, and I
couldn't find any lossy-webp-with-alpha files that use uncompressed
alpha channels. So the code here isn't really tested.
If someone comes along who wants to implement lossy webp decoding,
they now only need to implement decode_webp_chunk_VP8() and everything
might Just Work.
It also makes it possible to implement alpha chunk decoding before
implementing lossy decoding (by making decode_webp_chunk_VP8()
return an empty black bitmap for testing).
That way, animated and non-animated webp files use the same code path
to decode images. That will make it easier to add handling for lossy
decompression and for alpha chunk handling.
No behavior change.
"The official project language is American English […]."
5d2e915623/CONTRIBUTING.md (L30)
Here's a short statistic of the occurrences of the word "behavio(u)r":
$ git grep -IPioh 'behaviou?r' | sort | uniq -c | sort -n
2 BEHAVIOR
24 Behaviour
32 behaviour
407 Behavior
992 behavior
Therefore, it is clear that "behaviour" (56 occurrences) should be
regarded a typo, and "behavior" (1401 occurrences) should be preferred.
Note that The occurrences in LibJS are intentionally NOT changed,
because there are taken verbatim from the specification. Hence:
$ git grep -IPioh 'behaviou?r' | sort | uniq -c | sort -n
2 BEHAVIOR
10 behaviour
24 Behaviour
407 Behavior
1014 behavior
With this, lossless animated webp files work :^)
(Missing: Loop count handling is not yet implemented, and alpha blending
between frames isn't done in linear space.)
This also slightly refactors things to share more implementation with
the SVG linear gradients, and improve accuracy (which fixes some banding
issues).
LutAToBTagData::from_bytes() and LutBToATagData::from_bytes() already
reject curves for which this isn't true with an error.
Ensure potential future callers of the constructors get it right too.
This allows converting to a color space that uses a non-parametric
curve, for example:
Build/lagom/image -o foo.png \
--convert-to-color-profile .../profiles/sRGB-v2-micro.icc \
input.jpg
...where profiles/sRGB-v2-micro.icc is from
https://github.com/saucecontrol/Compact-ICC-Profiles/
(Parametric curves are new in ICC v4, which means all v2 profiles
use point curves.)
For now, only for color spaces that are supported by Profile::to_pcs()
and Profile::from_pcs(), which currently means that all matrix profiles
(but not LUT profiles) in the source color space work, and that
matrix profiles with parametric curves in the destination color
space work.
This adds Profile::convert_image(Bitmap, source_profile), and
adds a `--convert-to-color-profile file.icc` flag to `image`.
It only takes a file path, so to use it with the built-in
sRGB profile, you have to write it to a file first:
% Build/lagom/icc -n sRGB --reencode-to serenity-sRGB.icc
`image` by default writes the source image's color profile
to the output image, and most image viewers display images
looking at the profile.
For example, take `Seven_Coloured_Pencils_(rg-switch_sRGB).jpg`
from https://commons.wikimedia.org/wiki/User:Colin/BrowserTest.
It looks normal in image viewers because they apply the unusual
profile embedded in the profile. But if you run
% Build/lagom/image -o huh.png --strip-color-profile \
'Seven_Coloured_Pencils_(rg-switch_sRGB).jpeg'
and then look at huh.png, you can see how the image's colors
look like when interpreted as sRGB (which is the color space
PNG data is in if the PNG doesn't store an embedded profile).
If you now run
% Build/lagom/image -o wow.png \
--convert-to-color-profile serenity-sRGB.icc --strip-color-profile \
'Seven_Coloured_Pencils_(rg-switch_sRGB).jpeg'
this will convert that image to sRGB, but then not write
the profile to the output image (verify with `Build/lagom/icc wow.png`).
It will look correct in image viewers, since they display PNGs without
an embedded color profile as sRGB.
(This works because 'Seven_Coloured_Pencils_(rg-switch_sRGB).jpeg'
contains a matrix profile, and Serenity's built-in sRGB profile
uses a matrix profile with a parametric curve.)
This implements conversion from profile connection space to the
device-dependent color for matrix-based profiles.
It only does the inverse color transform but does not yet do the
inverse tone reproduction curve transform -- i.e. it doesn't
implement many cases (LUT transforms), and it does the one thing
it does implement incorrectly. But to vindicate the commit a bit,
it also does the incorrect thing very inefficiently.
