We don't support all parts of the font formats we assume as "supported"
in the CSS parser. For example, if an open type font has a CFF table, we
reject loading it. This meant that until now, when such an
unsupported-supported font url was first in the list of urls, we
couldn't load it at all, even when we would support a later url.
To resolve that, try loading all font urls one after each other, in case
we are not able to load the higher priority one.
This also resolves a FIXME related to spec compliant url prioritization.
Our CSS parser already filters and prioritizes font src urls in
compliance with the spec. However, we still had to resort to brittle
file extension matching, because some websites don't set the `format`
and if the first url in a src list happened to be one we don't support,
the font could not be loaded at all. This now is unnecessary because we
can try and discard the urls instead.
Use FlyString::from_deprecated_fly_string() in these instances instead
of FlyString::from_utf8(). As we convert to new FlyString/String we want
to be aware of these potential unnecessary allocations.
This relied on pulling the current realm from the main thread VM, which
requires an execution context to be on the VM's stack. This heavily
relied on the dummy execution context that is always on the stack, for
example, when parsing the UA style sheets where no JavaScript is
running.
This requires Parser to be movable, so we remove the `default`
destructors from Parser and TokenStream, and give them both move
constructors. Since TokenStream only holds a reference to its tokens,
(and it needs to, to avoid copying when given eg a function's contents,)
we add a manual move constructor for Parser which creates a new
TokenStream from the new Parser's tokens, and then manually copies the
old TokenStream's state.
Instead of constructing a Tokenizer and then calling parse() on it, we
now call `Tokenizer::tokenize(...)` directly. (Renamed from `parse()`
because this is a Tokenizer, not a Parser.)
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>>.
Like the name suggests this pointer type compares its pointees by value
rather than just by the pointer. This is needed for the defaulted
struct Properties equality operator.
This commit also contains a few changes to StyleValue such as replacing
the operator==()s with a .equals() again. This is done to avoid the new
reversed operator==()s ambiguity in C++20.
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.
This will make it easier to support both string types at the same time
while we convert code, and tracking down remaining uses.
One big exception is Value::to_string() in LibJS, where the name is
dictated by the ToString AO.
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 :^)
Prevent media query parser from falling back into
MediaQuery::create_not_all() for queries with unknown media type.
With this change following test works correctly:
http://wpt.live/css/css-conditional/at-media-001.html
This parses conic-gradient()s while also attempting to share the bulk
of the parsing code with linear-gradient()s. This is done by extracting
the <color-stop-list> parsing to a "fill in the blacks" generic
function. This is a little awkward but cuts down on a lot of copy
pasting of code.
This parses <position> according to the following grammer:
<position> = [
[ left | center | right ] || [ top | center | bottom ]
|
[ left | center | right | <length-percentage> ]
[ top | center | bottom | <length-percentage> ]?
|
[ [ left | right ] <length-percentage> ] &&
[ [ top | bottom ] <length-percentage> ]
]
The code is a little hairy and simply tries each alternative in turn,
manually checking the possible orders. There may be a better way to
represent this.
Refactor various classes in the GridTrackSize file for the incoming
named_tracks feature.
Previously the ExplicitTrackSizing had mixed responsiblities with the
newly-named GridRepeat class. This made it so it was not possible to
have multiple repeats within a single 'GridTrackSizeList' definition.
The MetaGridTrackSize class had both the responsibilities of being a
container for minmax values as well as for simple GridSizes. By uniting
the different possible values (repeat, minmax, default) into the
ExplicitGridTrack class are able to be more expressive as to the
different grid size modalities.
The GridTrackSizeList will be useful as compared to a
Vector<ExplicitGridTrack> since this way can keep track of the declared
line names. These same line names are able to be declared within the
values of a repeat function, hence the presence of a GridTrackSizeList
inside the GridRepeat class.
Add classes ExplicitTrackSizing and MetaGridTrackSize which will allow
for managing properties like auto-fill and minmax.
In the following CSS example there are 3 classes that will be used:
grid-template-column: repeat(auto-fill, minmax(50px, 1fr) 75px);
ExplicitTrackSizing - will contain the entire value. e.g.
repeat(auto-fill, minmax(50px, 1fr) 75px)
With a flag if it's a repeat, as well as references to the
MetaGridTrackSizes which is the next step down.
MetaGridTrackSize:
Contain the individual grid track sizes. Here there are two:
minmax(50px, 1fr) as well as 75px.
This way can keep track if it's a minmax function or not, and the
references to both GridTrackSizes in the case it is, or in just the one
if it is not.
GridTrackSize:
Is the most basic element, in this case there are three in total; two of
which are held by the first MetaGridTrackSize, and the third is held by
the second MetaGridTrackSize.
Examples: 50px, 1fr and 75px.
Add ability to use values passed to grid-template-columns and
grid-template-rows for CSS Grid layout within a repeat() function.
E.g. grid-template-columns: repeat(2, 50px); means to have two columns
of 50px width each.
These changes improve the parsing of the span property, including
handling negative values for the span (defaults to 1), as well as when
no number is passed (also defaults to 1).
Implement span correctly when indicated in the grid-column-start,
grid-row-start, etc. CSS properties. Previously it had been implemented
as if span was something that went alongside the position property, but
actually it seems like if you do 'span 3' in the grid-column-start
property, for example, this means it literally spans 3 blocks, and the
3 has nothing to do with position.
When matching selectors in HTML documents, we know that all the elements
have lowercase local names already (the parser makes sure of this.)
Style sheets still need to remember the original name strings, in case
we want to match against non-HTML content like XML/SVG. To make the
common HTML case faster, we now cache a lowercase version of the name
with each type/class/id SimpleSelector.
This makes tag type checks O(1) instead of O(n).
I don't see any indication in the spec that an empty prelude should be
disallowed. This fixes rules like `@font-face{...}` (note the absence
of whitespace before the `{`.)
To follow spec more closely, only set the has_span value if there is a
number immediately following, like "span 2". Otherwise the span value
should be ignored.
This is a monster patch that turns all EventTargets into GC-allocated
PlatformObjects. Their C++ wrapper classes are removed, and the LibJS
garbage collector is now responsible for their lifetimes.
There's a fair amount of hacks and band-aids in this patch, and we'll
have a lot of cleanup to do after this.
This commit fixes a bug found when passing exotic values in the
grid-template-columns (or grid-template-rows) which are not yet
supported.
The bug seems to have been something like:
grid-template-columns: 0 minmax(0, calc(10px - var(--some-color)));
Add functionality to begin parsing grid-template-columns and
grid-template-rows. There are still things to be added, like parsing
functions, but I would say a couple of the major points are already
adressed like length, percentage, and flexible-length.
The only accepted syntax for these seems to be
<color> <length percentage> <length percentage>, no other order.
But that's just gathered from looking at other browsers as though
these are supported by all major browsers, they don't appear in
the W3C spec.
The -webkit version of linear-gradient does not include the `to`
before a <side or corner>. The angles of the <side or corner>
for the webkit version are also opposite that of the standard one.
So for the standard: linear-gradient(to left, red, blue)
The webkit version is: -webkit-linear-gradient(right, red, blue)
Adding the `to` in the -webkit version is invalid, omitting it in
the standard one is also invalid.
These will be needed for styling progress bars, sadly this can
only be done with these non-standard selectors. These are, hovever,
in use on real sites such as https://rpcs3.net/compatibility.
As noted, we should probably handle calc() parsing as part of parsing
other values. eg, any `<length>` can be a `calc()` that returns a
length, but we currently have to manually handle that everywhere that
doesn't use the `Parser::parse_css_value(ComponentValue)` method.
This prevents us from needing a sv suffix, and potentially reduces the
need to run generic code for a single character (as contains,
starts_with, ends_with etc. for a char will be just a length and
equality check).
No functional changes.
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.
This commit has no behavior changes.
In particular, this does not fix any of the wrong uses of the previous
default parameter (which used to be 'false', meaning "only replace the
first occurence in the string"). It simply replaces the default uses by
String::replace(..., ReplaceMode::FirstOnly), leaving them incorrect.
The lack of the commit() before returning the x_value here meant,
that in parse_background_value() the token stream would be one token
behind after parsing the background-size. This led to it to returning
null, after it sees the unexpected 'second' contain / cover token.
With this change all of backgrounds.html is working again.
This prevents font-face rules without a block statement from crashing
LibWeb during CSS parsing.
The issue was discovered by Lubrsi during CSS parser fuzzing. :)
Fixes#14141.
Implement parsing of rgb(..) and hsl(..) in both the modern level 4
syntax without commas, and the legacy syntax with commas.
The parser accepts non-integer numbers but rounds to integer values
for now.
Previously, `var()` inside functions like `rgb()` wasn't resolved.
This will set the background color for badges in the New category on
https://ports.serenityos.net. :^)
When parsing <ndash-dimension> <signless-integer>, we tried to parse
a new token from the stream instead of using the value we had already
extracted. This caused pages that used the syntax to crash.
...and change how the two parsing steps fit together.
The two steps were previously quite muddled. Both worked with the
TokenStream directly, and both were responsible for rewinding that
stream if there was an error. This is both confusing and also made it
impossible to replace the rewinding with StateTransactions.
This commit more clearly divides the work between the two functions: One
parses ComponentValues and produces a string, and the other parses that
string to produce the UnicodeRange. It also replaces manual rewinding
in the former with StateTransactions.
This should be a bit easier to follow.
parse_media_query() no longer rewinds if the media query is invalid,
because it then interprets all the tokens as a "not all" query.
`a` and `b` had to be declared at the top of the function before since
they were used by the `make_return_value()` lambda. But now that
doesn't exist, we can move them to where they are used - or eliminate
them entirely.
parse_a_n_plus_b_pattern()'s job is to parse as much of the TokenStream
as it can as a An+B, and then stop. The caller can then deal with any
trailing tokens as it wishes.
...using a ParseErrorOr type alias.
This lets us replace a bunch of manual error-checking with TRY. :^)
I also replaced the ParsingResult::Done value with returning an
Optional. I wasn't happy with treating "Done" as an error when I first
wrote this, and this makes a clear distinction between the two.
The spec grammar for `text-decoration-line` is:
`none | [ underline || overline || line-through || blink ]`
Which means that it's either `none`, or any combination of the other
values. This patch makes that parse for `text-decoration-line` and
`text-decoration`, stores the results as a Vector, and adjusts
`paint_text_decoration()` to run as a loop over all the values that are
provided.
As noted, storing a Vector of values is a bit wasteful, as they could be
stored as flags in a single `u8`. But I was getting too confused trying
to do that in a nice way.
As before, this requires deviating from the spec slightly to create the
StyleRule fully-formed instead of creating it empty and then modifying
its internals.
This means deviating slightly from the spec in order to construct a
fully-initialized Declaration instead of creating an empty one and then
poking at its internals.
DeclarationOrAtRule should probably use a Variant, but for now, making
its Declaration member optional is quick and easy.
This means deviating a little from the spec, so that we create a
complete Block in one go instead of creating an empty one and then
poking at its internals.
The goal here is to move the parser-internal classes into this namespace
so they can have more convenient names without causing collisions. The
Parser itself won't collide, and would be more convenient to just
remain `CSS::Parser`, but having a namespace and a class with the same
name makes C++ unhappy.
This is used to skip downloading fonts in formats that we don't support.
Currently we only support TTF as far as I am aware.
The parts of a `src` are in a fixed order, unusually, which makes the
parsing more nesty instead of loopy.
Like, An+B, this is an old construct that does not fit well with modern
CSS syntax, so things get a bit hairy! We have to determine which
tokens match the grammar for `<urange>`, then turn those back into a
string, and then parse the string differently from normal. Thankfully
the spec describes in detail how to do that. :^)
This is not 100% correct, since we are not using the original source
text (referred to in the spec as the "representation") of the tokens,
but just converting them to strings in a manual, ad-hoc way.
Re-engineering the Tokenizer to keep that original text was too much of
a tangent for today. In any case, we do parse `U+4???`, `U+0-100`,
`U+1234`, and similar, so good enough for now!
"Component value" is the term used in the spec, and it doesn't conflict
with any other types, so let's use the shorter name. :^)
Also, this doesn't need to be friends with the Parser any more.
We're calling this in a way that is incorrect, and so the algorithm's
assumption that the next token is an `<ident-token>` is wrong, and we
have to handle that failing. Ideally we would just stop calling this
incorrectly, but until then, let's actually document what is happening.
The code had to change a bit to match. Previously, we appended an empty
sub-list immediately, but now we append it at the end. The difference
is that if there are no tokens, we now correctly return an empty
list-of-lists, instead of a list containing an empty list.
We now correctly call convert_to_rule() outside of this function.
As before, I've renamed `parse_as_rule()` -> `parse_as_css_rule()` to
match the free function that calls it.
This is not actually used by anything currently, but it should be used
for `@media` and other at-rules.
Removed the public parse_as_list_of_rules() because public functions
should be things that outside classes actually need to use.
`parse_a_stylesheet()` should not do any conversion on its rules. This
change corrects that. There are other places where we get this wrong,
but one thing at a time. :^)
CSS Values and Units Module Level 5 defines attr as:
`attr(<q-name> <attr-type>?, <declaration-value>?)`
This implementation does not contain support for the type argument,
effectively supporting `attr(<q-name>, <declaration-value>?)`
When parsing the "style" attribute on elements, we'd previously ask the
CSS parser for a PropertyOwningCSSStyleDeclaration. Then we'd create a
new ElementCSSInlineStyleDeclaration and transfer the properties from
the first object to the second object.
This patch teaches the parser to make ElementCSSInlineStyleDeclaration
objects directly.
When I wrote the An+B parser, it was guaranteed to have no
non-whitespace tokens after it. This is no longer true with the `of
foo` syntax, so this patch corrects the logic when there is no `+B`
segment.
This makes this case shown on Twitter work correctly. :^)
https://twitter.com/simevidas/status/1506657566012678151
`text-shadow` does not support this, so this way we can still use the
same parsing code.
It's OK that we still assign a ShadowPlacement value to the
ShadowStyleValue, since it will just get ignored when painting
text-shadows, but if it appears in the property value then that is a
syntax error.
The `text-shadow` property is almost identical to `box-shadow`:
> Values are interpreted as for box-shadow [CSS-BACKGROUNDS-3].
> (But note that the inset keyword are not allowed.)
So, let's use the same data structures and parsing code for both. :^)
There's really no reason to use doubles here, except at the time I
wanted to use doubles everywhere in CSS. I now realize that is
excessive, so everything can be floats instead.
There was no real benefit to creating the SimpleSelector early and then
modifying it, and doing so made this code harder to follow than it
needs to be.
The spec says:
> <delim-token> has a value composed of a single code point.
So using StringView is a bit overkill.
This also allows us to use switch statements in the future.
This doesn't have parsing support for multiple languages in the same
selector. Support for language subcodes is not great either. But it
does do the basics.
In Selectors level 4, `:nth-child()` and `:nth-last-child()` can both
optionally take a selector-list argument. This selector-list acts as a
filter, so that only elements matching the list are counted. For
example, this means that the following are equivalent:
```css
:nth-child(2n+1 of p) {}
p:nth-of-type(2n+1) {}
```
`<forgiving-selector-list>` and `<forgiving-relative-selector-list>` are
the same as regular selector-lists, except that an invalid selector
does not make the whole list invalid. The former is used by the `:is()`
pseudo-class.
For example:
```css
/* This entire selector-list is invalid */
.foo, .bar, !?invalid { }
/* This is valid, but the "!?invalid" selector is removed */
:is(.foo, .bar, !?invalid) { }
```
Also as part of this, I've removed the `parse_a_selector(TokenStream)`
and `parse_a_relative_selector(TokenStream)` methods as they don't add
anything useful.
This resolves the ambiguity between whether a single number is a number
or a ratio. :^)
Also removed the "no more tokens" checks from
deea129b8c - that logic was completely
wrong, since there are always tokens after a value in the `(123 < foo <
456)` syntax.
As noted, the Parser can't handle the `<number>` syntax for this - it
gets parsed instead by the `<number>` branch. We can't actually resolve
the ambiguity without making the Parser aware of what type each
media-feature is, but I will get to that soon. :^)
For now, we only understand `none`, `normal`, `<image>` and `<string>`.
The various other functions and identifiers can be added later.
We can *almost* use a StyleValueList for this, except it's divided into
two parts - the content, and the optional "alt text". So, I've added a
new StyleValue for it.
This is the only dimension type besides `<length>` that is used in any
media queries in levels 4 or 5 right now. Others can be included
if/when they're needed.
This means the units are defined in a single place instead of two.
Also removed the verify that we didn't produce a bogus % dimension token
in the Tokenizer, since this has never happened and the parser is not a
tokenizer test suite. :^)
When converting this code to use Optional, I accidentally left in the
initialization, so it *always* had a value, and always created a Length
from it. Oops.
The pattern we've adopted for other multi-value properties is to run in
a loop like this, since that makes it easier to cater for values
appearing in different orders.
During the LengthPercentage split, I converted the individual-corner
`border-foo-bar-radius` properties to LengthPercentage but forgot
`border-radius` itself! Oops. Discord's CSS was doing `border-radius:
50%` a lot, so this cuts down on CSS parser spam.
Browser has a handy debug menu option to dump all stylesheets, so we
don't need to spam the console with this. (All the spam massively slows
down page loads.)
See https://www.w3.org/TR/css-values-3/#calc-type-checking
If the sub-expressions' types are incompatible, we discard the calc() as
invalid.
Had to do some minor rearranging/renaming of the Calc structs to make
the `resolve_foo_type()` templates work too.
This lets us produce valid CSS in its to_string() method, instead of
always adding commas as before. :^)
Also, finally added a Formatter for StyleValues.
This is in a slightly weird state, where Percentages are sometimes
Lengths and sometimes not, which I will be cleaning up in subsequent
commits, in an attempt not to change all of LibWeb in one go. :^)
This means you can now do queries like:
```css
@media (400px <= width < 800px) { }
```
Chromium and Firefox which I tested with both don't support this yet, so
that's cool. :^)
Past me decided that the grammar was overly verbose and I could do it
better myself. Which seemed fine until the spec changed and I didn't
know how to integrate the changes. Lesson learned! :^)
Rather than have a function for every single part of the grammar, I have
written some as lambdas, and combned `<media-condition>` and
`<media-condition-without-or>` into one function. But otherwise it's
close to the spec, with comments listing the part of the grammar being
parsed, so hopefully it will be easier to make future adjustments!
This does not add any new functionality.
Web::CSS::MediaQuery::MediaFeature::Type was getting a bit ridiculous!
Also, this moves the detection of "min-" and "max-" media-features into
the MediaFeature itself, since this is an implementation detail, not
part of the spec.
Previously, we were using StyleValues for this, which was a bit of a
hack and was brittle, breaking when I modified how custom properties
were parsed. This is better and also lets us limit the kinds of value
that can be used here, to match the spec.
If a property is custom or contains a `var()` reference, it cannot be
parsed into a proper StyleValue immediately, so we store it as an
UnresolvedStyleValue until the property is compute. Then, at compute
time, we resolve them by expanding out any `var()` references, and
parsing the result.
The implementation here is very naive, and involves copying the
UnresolvedStyleValue's tree of StyleComponentValueRules while copying
the contents of any `var()`s it finds along the way. This is quite an
expensive operation to do every time that the style is computed.
Now that we can serialize CSS tokens, we can just hold a string and then
re-parse it when the Supports is evaluated. This feels a little weird,
but it only happens once so it's not going to slow it down much, and it
keep the API cleaner.
The changes in commit d462a6720a meant
that `CSSLoader::load_next_import_if_needed()` was no longer being
called, so no `@import` rules were loading.
The spec does not seem to mention when that loading should take place,
but WebKit does this at parse time, as each rule is added to the style
sheet. If it works for them, it's probably not that bad. :^)
So, that's what we now do here. The `CSSImportRule` creates a fetch
request when it is constructed, so each one is responsible for its own
contents.
Like the `background-foo` longhand properties (except
`background-color`), `background` allows multiple layers to be defined,
separated by commas. A layer does not necessarily contain something to
actually draw!
This parses as a `BackgroundStyleValue`, holding a `StyleValueList` for
each property. This is mostly to make expansion into longhands simpler -
if we had a list of `BackgroundStyleValue`s instead, one per layer, then
we would have to break it up per-property anyway when computing styles.
We now can parse lists of values for these properties:
- `background-attachment`
- `background-clip`
- `background-image`
- `background-origin`
- `background-position`
- `background-repeat`
- `background-size`
This uses two new Parser methods:
`parse_simple_comma_separated_value_list()` for the simple case when
each value is parsed from a single token; and
`parse_comma_separated_value_list()` which takes a lambda for when
parsing each value is more involved.
This also means that any unconsumed tokens at the end will make the
parsing fail as it should, where previously we just ignored them.
...as opposed to storing StyleValues, which we have to later check are
IdentifierStyleValues, which store identifiers that we can convert to
Repeat values later. It's fewer allocations, and we can't end up with
invalid values by mistake. :^)
This is step 1 in removing the two `background-repeat-x/y`
pseudo-properties. Since adding the concept of compound StyleValues, we
don't need `background-repeat` to be split in two any more.
This is done a bit differently from other properties: using a
TokenStream instead of just a Vector of ComponentValues. The reason for
this is, we can then use call the same function when parsing the
`background` shorthand. Otherwise, we would have to know in advance how
many values to pass down, which basically would involve duplicating the
`background-position` parsing code inside `background`.
The StyleValue is PositionStyleValue, since it represents a
`<position>`: https://www.w3.org/TR/css-values-4/#typedef-position
Unfortunately, background-position's parsing is a bit different from
`<position>`'s, (background-position allows 3-value syntax and
`<position>` doesn't) so we'll need to come back and write a different
parsing function for that later.
This accounts for cases like:
```css
.foo {
color: blue ! important ;
}
```
That's rare now that minifying is so popular, but does appear on some
websites.
I've added spec comments to `consume_a_declaration()` while I was at it.
Noticed this while checking some MediaWiki-based sites. It's not
obvious, but the spec does allow this, by not mentioning it in this list
of places whitespace is forbidden:
https://www.w3.org/TR/selectors-4/#white-space
The only reason it wasn't const before (and why we had a const_cast
hack) was to support ImageStyleValue's constructor taking it, which no
longer applies. `hack_count--;` :^)
This always felt awkward to me, and required a few other hacks to make
it work. Now, the request is only started when `load_bitmap()` is
called, which we do inside `NodeWithStyle::apply_style()`.
Previously, `box-shadow: none` would fail to parse, meaning that in this
example:
```css
p {
box-shadow: 20px 10px 5px magenta;
}
p.foo {
box-shadow: none;
}
```
... a `<p class="foo">` would still have a box-shadow, when it should
not have one. Now, we handle the `none` value. :^)
We can reduce the amount of padding the compiler adds in order to
ensure data alignment of member variables by ordering the types in
a struct by size in decending order.
Found By PVS-Studio: https://pvs-studio.com/en/docs/warnings/v802/
