Previously, if one operator returned an error, the TRY() would cause
us to return without restoring the outer graphics state, leading to
problems such as handing a 3-tuple to a grayscale color space
(because the inner object set up a grayscale color space that we
failed to dispose of).
Makes us crash later on page 43 of
https://devstreaming-cdn.apple.com/videos/wwdc/2017/821kjtggolzxsv/821/821_get_started_with_display_p3.pdf
Previously, every time a page switched fonts, we'd completely
re-parse the font.
Now, we cache fonts in Renderer, effectively caching them per page.
It'd be nice to have an LRU cache across pages too, but that's a
bigger change, and this already helps a lot.
Font size is part of the cache key, which means we re-parse the same
font at different font sizes. That could be better too, but again,
it's a big help as-is already.
Takes rendering the 1310 pages of the PDF 1.7 reference with
Build/lagom/bin/pdf --debugging-stats \
~/Downloads/pdf_reference_1-7.pdf
from 71 s to 11s :^)
Going through pages especially in the index is noticeably snappier.
(On the PDF 2.0 spec, ISO_32000-2-2020_sponsored.pdf, it's less
dramatic: From 19s to 16s.)
We now track it in the graphics state. It isn't used for anything yet.
Fixes the one thing that rendering the first 100 pages of
pdf_reference_1-7.pdf complains about.
This class had slightly confusing semantics and the added weirdness
doesn't seem worth it just so we can say "." instead of "->" when
iterating over a vector of NNRPs.
This patch replaces NonnullRefPtrVector<T> with Vector<NNRP<T>>.
The PDFFont class hierarchy was very simple (a top-level PDFFont class,
followed by all the children classes that derived directly from it).
While this design was good enough for some things, it didn't correctly
model the actual organization of font types:
* PDF fonts are first divided between "simple" and "composite" fonts.
The latter is the Type0 font, while the rest are all simple.
* PDF fonts yield a glyph per "character code". Simple fonts char codes
are always 1 byte long, while Type0 char codes are of variable size.
To this effect, this commit changes the hierarchy of Font classes,
introducing a new SimpleFont class, deriving from PDFFont, and acting as
the parent of Type1Font and TrueTypeFont, while Type0 still derives from
PDFFont directly. This distinction allows us now to:
* Model string rendering differently from simple and composite fonts:
PDFFont now offers a generic draw_string method that takes a whole
string to be rendered instead of a single char code. SimpleFont
implements this as a loop over individual bytes of the string, with
T1 and TT implementing draw_glyph for drawing a single char code.
* Some common fields between T1 and TT fonts now live under SimpleFont
instead of under PDFfont, where they previously resided.
* Some other interfaces specific to SimpleFont have been cleaned up,
with u16/u32 not appearing on these classes (or in PDFFont) anymore.
* Type0Font's rendering still remains unimplemented.
As part of this exercise I also took the chance to perform the following
cleanups and restructurings:
* Refactored the creation and initialisation of fonts. They are all
centrally created at PDFFont::create, with a virtual "initialize"
method that allows them to initialise their inner members in the
correct order (parent first, child later) after creation.
* Removed duplicated code.
* Cleaned up some public interfaces: receive const refs, removed
unnecessary ctro/dtors, etc.
* Slightly changed how Type1 and TrueType fonts are implemented: if
there's an embedded font that takes priority, otherwise we always
look for a replacement.
* This means we don't do anything special for the standard fonts. The
only behavior previously associated to standard fonts was choosing an
encoding, and even that was under questioning.
Errors can (and do) occur when trying to render text, and so far we've
silently ignored them, making us think that all is well when it isn't.
Letting show_text return errors will allow us to inform the user about
these errors instead of having to hiding them.
While the clipping logic was correct (current v/s new clipping path),
the clipping path contents weren't. This commit fixed that.
We calculate the clipping path in two places: when we set it to be the
whole page at graphics state creation time, and when we perform clipping
path intersection to calculate a new clipping path. The clipping path is
then used to limit painting by passing it to the painter (more
precisely, but passing its bounding box to the painter, as the latter
doesn't support arbitrary path clipping). For this last point the
clipping path must be in device coordinates.
There was however a mix of coordinate systems involved in the creation,
update and usage of the clipping path:
* The initial values of the path (i.e., the whole page) were in user
coordinates.
* Clipping path intersection was performed against m_current_path,
which is in device coordinates.
* To perform the clipping operation, the current clipping path was
assumed to be in user coordinates.
This mix resulted in the clipping not working correctly depending on the
zoom level at which one visualised a page.
This commit fixes the issue by always keeping track of the clipping path
in device coordinates. This means that the initial full-page contents
are now converted to device coordinates before putting them in the
graphics state, and that no mapping is performed when applied the
clipping to the painter.
When rendering text, a sequence of bytes corresponds to a glyph, but not
necessarily to a character. This misunderstanding permeated through the
Encoding through to the Font classes, which were all trying to calculate
such values. Moreover, this was done only to identify "space"
characters/glyphs, which were getting a special treatment (e.g., avoid
rendering). Spaces are not special though -- there might be fonts that
render something for them -- and thus should not be skipped
DeprecatedFlyString relies heavily on DeprecatedString's StringImpl, so
let's rename it to A) match the name of DeprecatedString, B) write a new
FlyString class that is tied to String.
Instead of calling TODO(), which will abort the program, we now return
an Error specifying that we haven't implemented the drawing operation
yet. This will now nicely trickle up all the way through to the
PDFViewer, which will then notify its clients about the problem.
The current rendering routine aborts as soon as an error is found during
rendering, which potentially severely limits the contents we show on
screen. Moreover, whenever an error happens the PDFViewer widget shows
an error dialog, and doesn't display the bitmap that has been painted so
far.
This commit improves the situation in both fronts, implementing
rendering now with a best-effort approach. Firstly, execution of
operations isn't halted after an operand results in an error, but
instead execution of all operations is always attempted, and all
collected errors are returned in bulk. Secondly, PDFViewer now always
displays the resulting bitmap, regardless of error being produced or
not. To communicate errors, an on_render_errors callback has been added
so clients can subscribe to these events and handle them as appropriate.
After adding support for XObject Form rendering, the next was to display
XObject images. This commit adds this initial support,
Images come in many shapes and forms: encodings: color spaces, bits per
component, width, height, etc. This initial support is constrained to
the color spaces we currently support, to images that use 8 bits per
component, to images that do *not* use the JPXDecode filter, and that
are not Masks. There are surely other constraints that aren't considered
in this initial support, so expect breakage here and there.
In addition to supporting images, we also support applying an alpha mask
(SMask) on them. Additionally, a new rendering preference allows to skip
image loading and rendering altogether, instead showing an empty
rectangle as a placeholder (useful for when actual images are not
supported). Since RenderingPreferences is becoming a bit more complex,
we add a hash option that will allow us to keep track of different
preferences (e.g., in a HashMap).
ColorSpaces can be specified in two ways: with a stream as operands of
the color space operations (CS/cs), or as a separate PDF object, which
is then referred to by other means (e.g., from Image XObjects and other
entities). These two modes of addressing a ColorSpace are slightly
different and need to be addressed separately. However, the current
implementation embedded the full logic of the first case in the routine
that created ColorSpace objects.
This commit refactors the creation of ColorSpace to support both cases.
First, a new ColorSpaceFamily class encapsulates the static aspects of a
family, like its name or whether color space construction never requires
parameters. Then we define the supported ColorSpaceFamily objects.
On top of this also sit a breakage on how ColorSpaces are created. Two
methods are now offered: one only providing construction of no-argument
color spaces (and thus taking a simple name), and another taking an
ArrayObject, hence used to create ColorSpaces requiring arguments.
Finally, on top of *that* two ways to get a color space in the Renderer
are made available: the first creates a ColorSpace with a name and a
Resources dictionary, and another takes an Object. These model the two
addressing modes described above.
We have a new, improved string type coming up in AK (OOM aware, no null
state), and while it's going to use UTF-8, the name UTF8String is a
mouthful - so let's free up the String name by renaming the existing
class.
Making the old one have an annoying name will hopefully also help with
quick adoption :^)
This implementation currently handles Form XObjects only, skipping
image XObjects. When rendering an XObject, its resources are passed to
the underlying operations so they use those instead of the Page's.
Operators usually assume that the resources its operations will require
will be the Page's. This assumption breaks however when XObjects with
their own resources come into the picture (and maybe other cases too).
In that case, the XObject's resources take precedence, but they should
also contain the Page's resources. Because of this, one can safely use
the XObject resources alone when given, and default to the Page's if
not.
This commit adds all operator calls an extra argument with optional
resources, which will be fed by XObjects as necessary.
Resources can come from other sources (e.g., XObjects), and since the
only attribute we are reading from Page are its resources it makes sense
to receive resources instead. That way we'll be able to pass down
arbitrary resources that are not necessarily declared at the page level.
Paths rendering was buggy because the map() function that translates
points from user space to bitmap space applied the vertical flip
conversion that the current transformation matrix already considers;
Hence, all paths were upside down. The only exception was the "re"
instruction, which manually adjusted the Y coordinate of its points to
be flipped again (and had a FIXME saying that this should be
unnecessary).
This commit fixes the map() function that maps userspace points to
bitmap coordinates. The "re" operator implementation has also been
simplified creating a rectangle first and mapping *that* instead of
mapping each point individually.
A new struct allows users to specify specific rendering preferences that
the Renderer class might use to paint some Document elements onto the
target bitmap. The first toggle allows rendering (or not) the clipping
paths on a page, which is useful for debugging.
The existing path clipping support was broken, as it performed the
clipping operation as soon as the path clipping commands (W/W*) were
received. The correct behavior is to keep a clipping path in the
graphic state, *intersect* that with the current path upon receiving
W/W*, and apply the clipping when performing painting operations. On top
of that, the intersection happening at W/W* time does not affect the
painting operation happening on the current on-build path, but takes
effect only after the current path is cleared; therefore a current and a
next clipping path need to be kept track of.
Path clipping is not yet supported on the Painter class, nor is path
intersection. We thus continue using the same simplified bounding box
approach to calculate clipping paths.
Since now we are dealing with more rectangles-as-path code, I've made
helper functions to build a rectangle path and reuse it as needed.
It was previously the job of the renderer to create fonts, load
replacements for the standard 14 fonts and to pass the font size back
to the PDFFont when asking for glyph widths.
Now, the renderer tells the font its size at creation, as it doesn't
change throughout the life of the font. The PDFFont itself is now
responsible to decide whether or not it needs to use a replacement
font, which still is Liberation Serif for now.
This means that we can now render embedded TrueType fonts as well :^)
It also makes the renderer's job much more simple and leads to a much
cleaner API design.
Previously we would draw all text, no matter what font type, as
Liberation Serif, which results in things like ugly character spacing.
We now have partial support for drawing Type 1 glyphs, which are part of
a PostScript font program. We completely ignore hinting for now, which
results in ugly looking characters at low resolutions, but gain support
for a large number of typefaces, including most of the default fonts
used in TeX.
As per spec, the positioning (or kerning) parameter of this operator
should translate the text matrix before the next showing of text.
Previously, this calculation was slightly wrong and also only applied
after the text was already shown.
Each of these strings would previously rely on StringView's char const*
constructor overload, which would call __builtin_strlen on the string.
Since we now have operator ""sv, we can replace these with much simpler
versions. This opens the door to being able to remove
StringView(char const*).
No functional changes.
