neurocyte/flow
neurocyte/flow
1
1
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

feat: improve performance of incremental re-parsing

Browse source 
We use Buffer.Node.get_byte_pos instead of Buffer.Node.get_range to calculate
tree-sitter edits for incremental parsing. This reduces the number of
egc_length calls required per syntax.edit call to the minimum.
This commit is contained in:
CJ van den Berg 2024-07-29 21:46:34 +02:00
parent b0a99a0813
commit 42ee6459e6
4 changed files with 106 additions and 19 deletions
Download patch file Download diff file Expand all files Collapse all files

54
src/buffer/Buffer.zig
View file

@ -585,9 +585,10 @@ const Node = union(enum) {
var ctx: Ctx = .{ .sel = sel, .out = copy_buf };
ctx.sel.normalize();
if (ctx.sel.begin.eql(ctx.sel.end))
return error.Stop;
return if (copy_buf) |_| "" else null;
self.walk_egc_forward(ctx.sel.begin.row, Ctx.walker, &ctx, metrics_) catch |e| return switch (e) {
error.NoSpaceLeft => error.NoSpaceLeft,
error.Stop => if (copy_buf) |buf_| buf_[0..ctx.bytes] else null,
else => error.Stop,
};
if (size) |p| p.* = ctx.bytes;
@ -599,7 +600,7 @@ const Node = union(enum) {
var end: Cursor = .{};
end.move_buffer_end(self, metrics);
const result = self.get_range(.{ .begin = start, .end = end }, result_buf, null, null, metrics) catch |e| switch (e) {
error.NoSpaceLeft => result_buf[0..],
error.NoSpaceLeft => result_buf,
else => @panic("buffer overflow in get_from_start_pos"),
};
return result orelse "";
@ -935,6 +936,55 @@ const Node = union(enum) {
return self.store(ctx.writer());
}
pub fn get_byte_pos(self: *const Node, pos_: Cursor, metrics_: Metrics) !usize {
const Ctx = struct {
line: usize = 0,
abs_col: usize = 0,
pos: Cursor,
byte_pos: usize = 0,
metrics: Metrics,
const Ctx = @This();
const Writer = std.io.Writer(*Ctx, error{Stop}, write);
fn write(ctx: *Ctx, bytes: []const u8) error{Stop}!usize {
if (ctx.line >= ctx.pos.row) {
return ctx.get_col_bytes(bytes, bytes.len);
} else for (bytes, 1..) |char, i| {
ctx.byte_pos += 1;
if (char == '\n') {
ctx.line += 1;
if (ctx.line >= ctx.pos.row)
return ctx.get_col_bytes(bytes[i..], bytes.len);
}
}
return bytes.len;
}
fn get_col_bytes(ctx: *Ctx, bytes: []const u8, result: usize) error{Stop}!usize {
var buf: []const u8 = bytes;
while (buf.len > 0) {
if (ctx.abs_col >= ctx.pos.col) return error.Stop;
if (buf[0] == '\n') return error.Stop;
var cols: c_int = undefined;
const egc_bytes = ctx.metrics.egc_length(ctx.metrics.ctx, buf, &cols, ctx.abs_col);
ctx.abs_col += @intCast(cols);
ctx.byte_pos += egc_bytes;
buf = buf[egc_bytes..];
}
return result;
}
fn writer(ctx: *Ctx) Writer {
return .{ .context = ctx };
}
};
var ctx: Ctx = .{
.pos = pos_,
.metrics = metrics_,
};
self.store(ctx.writer()) catch |e| switch (e) {
error.Stop => return ctx.byte_pos,
};
return error.NotFound;
}
pub fn debug_render_chunks(self: *const Node, line: usize, output: *ArrayList(u8), metrics_: Metrics) !void {
const ctx_ = struct {
l: *ArrayList(u8),