Type: refuse negative or fractional byte offsets in text commands

text.startList read a negative or fractional `at` as the end of the
text, so it could remove a typed list marker there and start a list.
The `start` and `end` of text.editRange, text.getRange,
text.setRangeStyle and the type shortcut commands given an `id` read
such values as the start or end of the text.

These now return "`at` must be a whole number" (naming the key) and
change nothing. Omitted or null keys keep their defaults, and 2.0 is
read as 2.
This commit is contained in:
dotCooCoo
2026-11-04 16:55:22 +07:00
committed by bflatastic
parent 1d459796ad
commit 595db10900
3 changed files with 47 additions and 6 deletions
+15 -6
View File
@@ -7,6 +7,7 @@ use vectorcraft_doc::{Appearance, CharStyle, Node, NodeId, NodeKind, TextKind, T
use vectorcraft_geom::{Affine, Point, Vec2};
use vectorcraft_text::edit;
use super::menucmds::whole_or;
use super::stroke::StrokeChange;
use super::typecmd::refresh_bounds;
use super::{
@@ -233,12 +234,21 @@ fn byte_param(p: &Value, k: &str) -> Option<usize> {
p.get(k).and_then(Value::as_u64).map(index)
}
/// The byte offset that `p` gives for `k`, or `default` when `p` leaves it out or gives null. Any
/// other value that is not a whole number of at least 0, such as `"2"`, `-1` or `1.5`, is an error.
fn offset_or(p: &Value, k: &str, default: usize, cmd: &str) -> Result<usize> {
match p.get(k) {
None | Some(Value::Null) => Ok(default),
Some(_) => Ok(index(whole_or(p, k, 0, cmd)?)),
}
}
/// The bytes `start..end` of `p` in text of `len` bytes, clamped and ordered (without them, all of
/// it). A `start` or `end` that is not a number or null, such as `"2"`, is an error.
/// it). A `start` or `end` that is not a whole number of at least 0 or null, such as `"2"`, `-1`
/// or `1.5`, is an error.
fn range_of(p: &Value, len: usize, cmd: &str) -> Result<(usize, usize)> {
check_numbers(p, &["start", "end"], cmd)?;
let a = byte_param(p, "start").unwrap_or(0).min(len);
let b = byte_param(p, "end").unwrap_or(len).min(len);
let a = offset_or(p, "start", 0, cmd)?.min(len);
let b = offset_or(p, "end", len, cmd)?.min(len);
Ok((a.min(b), a.max(b)))
}
@@ -310,10 +320,9 @@ fn clear_param(s: &Session, p: &Value, id: NodeId, c: &str) -> Result<Vec<NodeId
fn start_list(s: &mut Session, p: &Value) -> Result<Value> {
const C: &str = "text.startList";
let id = id_param(p, "id").ok_or_else(|| bad(C, "missing `id`"))?;
check_numbers(p, &["at"], C)?;
let t = text_ref(s, id)?;
let len = edit::runs_len(&t.runs);
let at = byte_param(p, "at").unwrap_or(len).min(len);
let at = offset_or(p, "at", len, C)?.min(len);
let Some((start, list)) = t.typed_list_at(at) else {
return Ok(json!({"id": id.0, "started": false, "caret": at}));
};
+27
View File
@@ -1200,6 +1200,33 @@ fn start_list_command_removes_the_typed_characters_and_starts_the_list() {
assert!(s.execute("text.startList", &json!({"at": 3})).is_err(), "id required");
}
/// A byte offset given as a negative or fractional number is an error, not the end of the text
/// (`at`) or its start or end (`start`, `end`), and the text is left as it was. Null keeps the default.
#[test]
fn byte_offsets_that_are_not_whole_numbers_are_refused() {
let mut s = session();
let id = text(&mut s, "x\n# ");
for at in [json!(-1), json!(1.5)] {
let e = s.execute("text.startList", &json!({"id": id.0, "at": at})).unwrap_err().to_string();
assert!(e.contains("`at`"), "{e}");
}
assert_eq!(obj(&s, id).plain_text(), "x\n# ", "no list was started");
assert_eq!(s.execute("text.startList", &json!({"id": id.0, "at": null})).unwrap()["started"], json!(true));
let id = text(&mut s, "Hello");
for (cmd, p, key) in [
("text.getRange", json!({"id": id.0, "start": -1, "end": 2}), "start"),
("text.getRange", json!({"id": id.0, "start": 1, "end": 2.5}), "end"),
("text.editRange", json!({"id": id.0, "start": 1.5, "end": 2, "insert": "!"}), "start"),
("text.editRange", json!({"id": id.0, "start": 1, "end": -2, "insert": "!"}), "end"),
("text.setRangeStyle", json!({"id": id.0, "start": 0.5, "size": 30}), "start"),
] {
let e = s.execute(cmd, &p).unwrap_err().to_string();
assert!(e.contains(&format!("`{key}`")), "{cmd} {p}: {e}");
}
assert_eq!(obj(&s, id).plain_text(), "Hello");
assert_eq!(s.execute("text.getRange", &json!({"id": id.0, "start": 1.0, "end": null})).unwrap()["text"], json!("ello"));
}
/// #1881: a Cmd/Ctrl drag while typing (the Selection tool lent to the Type tool) moves the type
/// and keeps what was typed, and typing goes on after it.
#[test]
+5
View File
@@ -1207,6 +1207,11 @@ and `amount` in `color.harmony`; `angle` in `paint.editGradient` and in the `gra
`width` and `tolerance` in `livePaint.strokeEdge`. A null is read as a key left out. A gradient stop that names a
swatch refuses a `tint` that is not a number with `` stop 0: `tint` must be a number ``.
A byte offset must be a whole number of at least 0. `at` in `text.startList`, and `start` and `end` in
`text.editRange`, `text.getRange`, `text.setRangeStyle` and the `type.step`, `type.size.*`, `type.resetTracking` and
`type.resetHorizontalScale` commands given an `id`, refuse a negative or fractional number with
`` `at` must be a whole number ``, naming the key, and change nothing. `2.0` is read as `2`, and a null as the key left out.
A step can use an earlier step's result, in `run` as in the `command_batch` tool. A parameter string that is exactly
`$N` or `$N.path` takes that value from the result of step `N` (counting from 0, in the order they ran; `run --in`'s
open isn't a step), and `$last…` from the step before. A path goes into objects by key and arrays by index: