This is a big one, I kind of over scoped a touch. I don't have it in me
to do a clean-up pass on this code before PR-ing it, so apologies for
any weirdness; I'll happily resolve any feedback, but trying to review
my own code right now is not something my brain wants to do.
#### Summary
- Font metric calculations reworked to be based primarily on data from
font tables, and now the majority of the logic is shared between
CoreText and FreeType, and we have more control over the specifics.
- Sprite font metrics for the position and thickness of underlines,
strikethroughs, and overlines are now separate, as well as box drawing
thickness, allowing for individual adjustments.
- Minimums are applied to font metrics to avoid *completely* broken
results from fonts with degenerate values in them.
### Unified Metrics Calculations
I added a solid foundation for parsing as many SFNT font tables as we
need, and added parsing for `head`, `hhea` (unused), `post`, and `OS/2`.
I didn't make a strong effort to account for variable fonts, so if a
variable font changes vertical metrics significantly it could cause
issues -- luckily, width, which is the most likely metric to change, is
not a problem, since we get the width by having our backends measure the
advances for us, which does account for variations correctly.
### Separated Sprite Metrics
I reworked the sprite renderer to just get a copy of the metrics that it
can use directly, instead of being given its own metrics (width, height,
thickness) -- so that the different thickness metrics can be used for
their intended purposes, and so that offsets for "unadjusted" characters
can be handled in the Box renderer itself.
### Minimums
Prevent degenerate fonts and bad `adjust` configs from creating, e.g.
zero-thickness underlines by applying a minimum to certain fields after
calculating metrics and after applying modifiers.
### Misc.
Fixed `init` for CoreText faces selecting the first predefined instance
of a variable font rather than the default.
This fixes the #2847. There were two problems:
1. Documentation made it easy to use the bash example as a template for
calling other integrations
2. f the ZSH integration script is called the same way as for bash, it
would clobber the options
This PR fixes both. Now the ZSH script can be simply called with
`source`.
Here is why the options were clobbered, and how the method of calling
affected it:
The `-L` flag (local scope) ensures that the emulation mode only applies
to the current function scope. When the function ends, the shell reverts
to its previous state. When called outside of a function, the flags
persist even after the script execution ends.
The recommended way of calling the ZSH integration included `autoload
-Uz` that created a function with the integration script as its body.
When called directly, the assumptions about being in the function body
broke.
This PR moves a lot of code into a function, so it's best to review with
whitespace ignored in the diff.
In multiple tests we create 1 or more pages by growing them 1 row at a
time, which results in an integrity check of the page for each row grown
which is just... horrible. By simply pausing integrity checks while
growing these pages (since growing them is not the point of the test) we
MASSIVELY speed up all of these tests.
Also reduced grapheme bytes during testing and made the Terminal "glitch
text" test actually assert what it intends to achieve, rather than just
blindly assuming 100 copies of the text will be the right amount -- this
lets us stop a lot earlier, making it take practically no time.
In multiple tests we create 1 or more pages by growing them 1 row at a
time, which results in an integrity check of the page for each row grown
which is just... horrible. By simply pausing integrity checks while
growing these pages (since growing them is not the point of the test) we
MASSIVELY speed up all of these tests.
Also reduced grapheme bytes during testing and made the Terminal "glitch
text" test actually assert what it intends to achieve, rather than just
blindly assuming 100 copies of the text will be the right amount -- this
lets us stop a lot earlier, making it take practically no time.
… break
Fixes#2941
This fixes the rendering of the text below. For those that can't see it,
it is the following in UTF-32: `0x22 0x1F3FF 0x22`.
```
"🏿"
```
`0x1F3FF` is the Fitzpatrick modifier for dark skin tone. It has the
Unicode property `Emoji_Modifier`. Emoji modifiers are defined in UTS
#51 and are only valid based on ED-13:
```
emoji_modifier_sequence := emoji_modifier_base emoji_modifier
emoji_modifier_base := \p{Emoji_Modifier_Base}
emoji_modifier := \p{Emoji_Modifier}
```
Additional quote from UTS #51:
> To have an effect on an emoji, an emoji modifier must immediately
follow
> that base emoji character. Emoji presentation selectors are neither
needed
> nor recommended for emoji characters when they are followed by emoji
> modifiers, and should not be used in newly generated emoji modifier
> sequences; the emoji modifier automatically implies the emoji
presentation
> style.
Our precomputed grapheme break table was mistakingly not following this
rule. This commit fixes that by adding a check for that every
`Emoji_Modifier` character must be preceded by an `Emoji_Modifier_Base`.
This only has a cost during compilation (table generation). The runtime
cost is identical; the table size didn't increase since we had leftover
bits we could use.
Fixes#2941
This fixes the rendering of the text below. For those that can't see it,
it is the following in UTF-32: `0x22 0x1F3FF 0x22`.
```
"🏿"
```
`0x1F3FF` is the Fitzpatrick modifier for dark skin tone. It has the
Unicode property `Emoji_Modifier`. Emoji modifiers are defined in UTS
#51 and are only valid based on ED-13:
```
emoji_modifier_sequence := emoji_modifier_base emoji_modifier
emoji_modifier_base := \p{Emoji_Modifier_Base}
emoji_modifier := \p{Emoji_Modifier}
```
Additional quote from UTS #51:
> To have an effect on an emoji, an emoji modifier must immediately follow
> that base emoji character. Emoji presentation selectors are neither needed
> nor recommended for emoji characters when they are followed by emoji
> modifiers, and should not be used in newly generated emoji modifier
> sequences; the emoji modifier automatically implies the emoji presentation
> style.
Our precomputed grapheme break table was mistakingly not following this
rule. This commit fixes that by adding a check for that every
`Emoji_Modifier` character must be preceded by an `Emoji_Modifier_Base`.
This only has a cost during compilation (table generation). The runtime
cost is identical; the table size didn't increase since we had leftover
bits we could use.
Load and display (`T`) was responding even with implicit IDs, because
the display was achieved with an early return in the transmit function
that bypassed the logic to silence implicit ID responses- by making it
not an early return we fix this.
This backs out commit bb185cf6b695420ce8b43b5c1cadd16ef71c481a.
This was breaking IME input for some users and overall I couldn't find
other users where this really fixed anything other than me so I'm going
to back this out and fix this using my own system.
Variable font init used to just select the first available predefined
instance, if there were any, which is often not desirable- using
createFontDescriptorFromData instead of createFontDescritorsFromData
ensures that the default variation config is selected. In the future we
should probably allow selection of predefined instances, but for now
this is the correct behavior.
I found this bug when adding the metrics calculation test case for
CoreText, hence why fixing it is part of the same commit.
Unify grid metrics calculations by relying on shared logic mostly based
on values directly from the font tables, this deduplicates a lot of code
and gives us more control over how we interpret various metrics.
Also separate metrics for underlined, strikethrough, and overline
thickness and position, and box drawing thickness, so that they can
individually be adjusted as the user desires.
The surface might be mutated during the clipboard confirmation (resized
in my case), leading to the copied cursor `page_pin` being invalidated.
Fixes#1714. Would be nice if @stgarf can verify this.
I agree to the MIT relicensing.
**Context**
Currently, if there are multiple keybindings with a shared prefix,
they are grouped into a nested series of Binding.Sets.
For example, as reported in #2734, the following bindings:
keybind = ctrl+z>1=goto_tab:1
keybind = ctrl+z>2=goto_tab:2
keybind = ctrl+z>3=goto_tab:3
Result in roughly the following structure (in pseudo-code):
Keybinds{
Trigger("ctrl+z"): Value.leader{
Trigger("1"): Value.leaf{action: "goto_tab:1"},
Trigger("2"): Value.leaf{action: "goto_tab:2"},
Trigger("3"): Value.leaf{action: "goto_tab:3"},
}
}
When this is formatted into a string (and therefore in +list-keybinds),
it is turned into the following as Value.format just concatenates
all the sibling bindings ('1', '2', '3') into consecutive bindings,
and this is then fed into a single configuration entry:
keybind = ctrl+z>1=goto_tab:1>3=goto_tab:3>2=goto_tab:2
**Fix**
To fix this, Value needs to produce a separate configuration entry
for each sibling binding in the Value.leader case.
So we can't produce the entry (formatter.formatEntry) in Keybinds
and need to pass information down the Value tree to the leaf nodes,
each of which will produce a separate entry with that function.
This is accomplished with the help of a new Value.formatEntries method
that recursively builds up the prefix for the keybinding,
finally flushing it to the formatter when it reaches a leaf node.
This is done without extra allocations by using a FixedBufferStream
with the same buffer as before, sharing it between calls to nested
siblings of the same prefix.
**Testing**
Besides the included unit tests, I ran the GLFW-based app
and verified that the resulting binary produced the correct output
with `ghostty +show-config`:
```
❯ .zig-cache//o/02a32e7ba516d2692577a46f1a0df682/ghostty +show-config 2>/dev/null | grep goto_tab
keybind = ctrl+z>1=goto_tab:1
keybind = ctrl+z>3=goto_tab:3
keybind = ctrl+z>2=goto_tab:2
```
**Caveats**
We do not track the order in which the bindings were added
so the order is not retained in the formatConfig output.
Resolves#2734
A common issue for US-centric users of a terminal is that the "option"
key on macOS is not treated as the "alt" key in the terminal.
## Background
macOS does not have an "alt" key, but instead has an "option" key. The "option"
key is used for a variety of purposes, but the troublesome behavior for some
(and expected/desired behavior for others) is that it is used to input special
characters.
For example, on a US standard layout, `option-b` inputs `∫`. This is not
a typically desired character when using a terminal and most users will
instead expect that `option-b` maps to `alt-b` for keybinding purposes
with whatever shell, TUI, editor, etc. they're using.
On non-US layouts, the "option" key is a critical modifier key for
inputting certain characters in the same way "shift" is a critical
modifier key for inputting certain characters on US layouts.
We previously tried to change the default for `macos-option-as-alt`
to `left` (so that the left option key behaves as alt) because I had the
wrong assumption that international users always used the right option
key with terminals or were used to this. But very quickly beta users
with different layouts (such as German, I believe) noted that this is
not the case and broke their idiomatic input behavior. This behavior was
therefore reverted.
## Solution
This confusing behavior happened frequently enough that I decided to
implement the more complex behavior in this commit. The new behavior is
that when a US layout is active, `macos-option-as-alt` defaults to true
if it is unset. When a non-US layout is active, `macos-option-as-alt`
defaults to false if it is unset. This happens live as users change
their keyboard layout.
**An important goal of Ghostty is to have zero-config defaults** that
satisfy the majority of users. Fiddling with configurations is -- for
most -- an annoying task and software that works well enough out of the
box is delightful. Based on surveying beta users, I believe this commit
will result in less configuration for the majority of users.
## Other Terminals
This behavior is unique amongst terminals as far as I know.
Terminal.app, Kitty, iTerm2, Alacritty (I stopped checking there) all
default to the default macOS behavior (option is option and special
characters are inputted).
All of the aforementioned terminals have a setting to change this
behavior, identical to Ghostty (or, Ghostty identical to them perhaps
since they all predate Ghostty).
I couldn't find any history where users requested the behavior of
defaulting this to something else for US based keyboards. That's
interesting since this has come up so frequently during the Ghostty
beta!
**Context**
Currently, if there are multiple keybindings with a shared prefix,
they are grouped into a nested series of Binding.Sets.
For example, as reported in #2734, the following bindings:
keybind = ctrl+z>1=goto_tab:1
keybind = ctrl+z>2=goto_tab:2
keybind = ctrl+z>3=goto_tab:3
Result in roughly the following structure (in pseudo-code):
Keybinds{
Trigger("ctrl+z"): Value.leader{
Trigger("1"): Value.leaf{action: "goto_tab:1"}),
Trigger("2"): Value.leaf{action: "goto_tab:2"}),
Trigger("3"): Value.leaf{action: "goto_tab:3"}),
}
}
When this is formatted into a string (and therefore in +list-keybinds),
it is turned into the following as Value.format just concatenates
all the sibling bindings ('1', '2', '3') into consecutive bindings,
and this is then fed into a single configuration entry:
keybind = ctrl+z>1=goto_tab:1>3=goto_tab:3>2=goto_tab:2
**Fix**
To fix this, Value needs to produce a separate configuration entry
for each sibling binding in the Value.leader case.
So we can't produce the entry (formatter.formatEntry) in Keybinds
and need to pass information down the Value tree to the leaf nodes,
each of which will produce a separate entry with that function.
This is accomplished with the help of a new Value.formatEntries method
that recursively builds up the prefix for the keybinding,
finally flushing it to the formatter when it reaches a leaf node.
This is done without extra allocations by using a FixedBufferStream
with the same buffer as before, sharing it between calls to nested
siblings of the same prefix.
**Caveats**
We do not track the order in which the bindings were added
so the order is not retained in the formatConfig output.
Resolves#2734
If a transmit and display command does not specify an ID or a number,
then it should not be responded to. We currently automatically assign
IDs in this situation, which isn't ideal since collisions can happen
which shouldn't, but this at least fixes the glaring observable issue
where transmit-and-display commands yield unexpected OK responses.
