pauljd/vectorcraft
1
0
Fork 0
mirror of https://github.com/storytold/vectorcraft.git synced 2026-11-13 21:17:48 +00:00
Code Issues Packages Projects Releases Wiki Activity

Scripts: E4X operations count their temporary storage against the heap limit

Browse Source 
Assigning to a named child counts its list of the matching children, and
assigning through an empty XMLList counts the chain of targets it climbs,
before either changes the tree. copy(), toXMLString(), normalize(), the
`..` selector and the depth check on insertion hold the stack they keep
for each open level of the tree. Inserting an XMLList, and assigning one
to an item of an XMLList, hold their copy of the list's items.

copy() reads attributes in place, and toXMLString() reads attributes and
namespace declarations in place, instead of copying the lists. copy()
charges the declarations it copies before it copies them. Deleting by
name or assigning an attribute compacts the remaining links in place
instead of building a second list.

toXMLString() holds the set of prefixes a namespace lookup passes, and
picks a new nsN prefix in one pass over the declarations in scope
instead of rescanning them for each candidate.

The parser holds its open elements and their names, the text it gathers,
each tag's attribute values, the sets that check declarations and
attribute names, and its one copy of an element's namespaces, and it
charges the strings of the namespace declarations it makes. The
allowance it holds from the source's length covers the copy a value
takes as it becomes a string.

Comparing XML values, converting them to strings, assigning an XMLList
to an attribute, and reading a name, a qualifier, a namespace URI or
embedded text from a string built by joining strings flatten it with
flat_charged. An attribute name read from a string reserves its copy,
and an element name shares the string.

Docs: docs/target-app-parity.md.
This commit is contained in:
dotCooCoo
2026-11-13 06:54:23 -08:00
parent e8c7421a3d
commit b1c684e912
3 changed files with 549 additions and 168 deletions
Download Patch File Download Diff File Expand all files Collapse all files
crates/js/src/xml.rs
+303 -168
View File
@@ -7,6 +7,7 @@
//! each new node, name, reference or output unit is charged to the heap before it is made.
use std::collections::{HashMap, HashSet};
use std::hash::Hash;
use std::rc::Rc;
use crate::LimitKind;
@@ -18,7 +19,7 @@ use crate::heap::{
XmlListData, XmlName, XmlNode, XmlNs, key, string_bytes,
};
use crate::interp::{Abrupt, ErrorType, HeldBytes, Interp, JsResult};
use crate::string_builtins::{UNITS_PER_FUEL, flat_charged};
use crate::string_builtins::{UNITS_PER_FUEL, flat_charged, push_held, reserve_units};
use crate::value::{Str, Value, index_key};
fn parser_units(s: &str) -> Vec<u16> {
@@ -295,26 +296,40 @@ fn name_bytes(n: &XmlName) -> usize {
string_bytes(n.local.len()) + n.uri.as_ref().map_or(0, |u| string_bytes(u.len())) + n.prefix.as_ref().map_or(0, |p| string_bytes(p.len()))
}
/// The namespace `stack` declares for `uri`, nearest first, skipping prefixes redeclared nearer.
/// `prefer` picks that prefix when several match; `need_prefix` rejects the default namespace.
fn ns_lookup(stack: &[XmlNs], uri: &Key, prefer: Option<&Key>, need_prefix: bool) -> Option<XmlNs> {
let mut seen: HashSet<&[u16]> = HashSet::new();
let mut best = None;
for ns in stack.iter().rev() {
let Some(p) = &ns.prefix else { continue };
if !seen.insert(p) {
continue;
}
if ns.uri == *uri && !(need_prefix && p.is_empty()) {
if prefer == Some(p) {
return Some(ns.clone());
}
if best.is_none() {
best = Some(ns.clone());
}
}
/// Inserts `item` into `set`, first holding an upper estimate of the bytes of the set's new table
/// when it must grow. Returns whether `item` was new.
fn insert_held<T: Eq + Hash>(interp: &mut Interp, held: &mut HeldBytes, set: &mut HashSet<T>, item: T) -> JsResult<bool> {
if set.len() == set.capacity() && !set.contains(&item) {
let more = set.capacity().max(4);
held.add(interp, more.saturating_mul(size_of::<T>() + 1).saturating_mul(16) / 7)?;
set.reserve(more);
}
best
Ok(set.insert(item))
}
/// Makes room in `list` for `more` items, first holding the bytes of its new capacity when it must
/// grow. The capacity at least doubles, as a push's would.
fn reserve_held<T>(interp: &mut Interp, held: &mut HeldBytes, list: &mut Vec<T>, more: usize) -> JsResult<()> {
if list.capacity() - list.len() < more {
let cap = list.len().saturating_add(more).max(list.capacity().saturating_mul(2));
held.add(interp, (cap - list.capacity()).saturating_mul(size_of::<T>()))?;
list.reserve_exact(cap - list.len());
}
Ok(())
}
/// The number `n` of a prefix written `ns<n>` in decimal without leading zeros, as
/// [`Interp::fresh_prefix`] makes them.
fn prefix_number(prefix: &[u16]) -> Option<usize> {
let digits = prefix.strip_prefix(&[u16::from(b'n'), u16::from(b's')])?;
let (&first, _) = digits.split_first()?;
if digits.len() > 9 || (first == u16::from(b'0') && digits.len() > 1) {
return None;
}
digits.iter().try_fold(0usize, |n, &c| {
let d = c.checked_sub(u16::from(b'0')).filter(|d| *d < 10)?;
Some(n * 10 + usize::from(d))
})
}
fn prefix_in_scope<'a>(stack: &'a [XmlNs], prefix: &[u16]) -> Option<&'a XmlNs> {
@@ -345,18 +360,6 @@ fn units_are(units: &[u16], text: &str) -> bool {
units.iter().copied().eq(text.encode_utf16())
}
/// A prefix no entry of `stack` uses.
fn fresh_prefix(stack: &[XmlNs]) -> Key {
let mut n = 0u32;
loop {
let candidate = key(&format!("ns{n}"));
if prefix_in_scope(stack, &candidate).is_none() {
return candidate;
}
n += 1;
}
}
/// One open element while serializing.
struct OutFrame {
node: ObjId,
@@ -579,13 +582,16 @@ impl Interp {
// ---- names ----
fn name_from_string(&mut self, s: &[u16]) -> JsResult<NameRef> {
/// The name `s` gives: an attribute name after `@`, which is copied after its units are
/// reserved, and otherwise an element name that shares `s`.
fn name_from_string(&mut self, s: &Key) -> JsResult<NameRef> {
if let Some((&0x40, rest)) = s.split_first() {
let uri = if rest == STAR { None } else { Some(Rc::from(Vec::new())) };
reserve_units(self, rest.len())?;
return Ok(NameRef { uri, local: Rc::from(rest), attribute: true });
}
let uri = if s == STAR { None } else { Some(self.xml_default_uri()?) };
Ok(NameRef { uri, local: Rc::from(s), attribute: false })
let uri = if **s == *STAR { None } else { Some(self.xml_default_uri()?) };
Ok(NameRef { uri, local: s.clone(), attribute: false })
}
/// ToXMLName, with a canonical array index read as a position (ECMA-357 section 10.6).
@@ -593,7 +599,7 @@ impl Interp {
match k {
Value::Number(n) if *n >= 0.0 && n.fract() == 0.0 && *n < 4_294_967_295.0 => Ok(Access::Index(*n as u32)),
Value::String(s) => {
let flat = s.flat();
let flat = flat_charged(self, s)?;
if let Some(i) = canonical_index(&flat) {
return Ok(Access::Index(i));
}
@@ -603,7 +609,8 @@ impl Interp {
if let Some(Class::QName { name, attribute, .. }) = self.heap.object(*o).map(|ob| &ob.class) {
return Ok(Access::Name(NameRef { uri: name.uri.clone(), local: name.local.clone(), attribute: *attribute }));
}
let s = self.to_str(k)?.flat();
let s = self.to_str(k)?;
let s = flat_charged(self, &s)?;
if let Some(i) = canonical_index(&s) {
return Ok(Access::Index(i));
}
@@ -787,31 +794,33 @@ impl Interp {
if n.attribute {
let count = self.xml_node(x)?.attrs.len();
self.xml_take_fuel(count)?;
let attrs = std::mem::take(&mut self.xml_node_mut(x)?.attrs);
let mut kept = Vec::with_capacity(attrs.len());
for a in attrs {
let mut kept = 0;
for k in 0..count {
let Some(a) = self.xml_node(x)?.attrs.get(k).copied() else { break };
let hit = self.xml_node(a)?.name.as_ref().is_some_and(|q| n.matches(q));
if hit {
self.xml_node_mut(a)?.parent = None;
} else {
kept.push(a);
self.xml_node_mut(x)?.attrs.swap(kept, k);
kept += 1;
}
}
self.xml_node_mut(x)?.attrs = kept;
self.xml_node_mut(x)?.attrs.truncate(kept);
return Ok(());
}
let count = self.children_len(x)?;
self.xml_take_fuel(count)?;
let children = std::mem::take(&mut self.xml_node_mut(x)?.children);
let mut kept = Vec::with_capacity(children.len());
for c in children {
let mut kept = 0;
for k in 0..count {
let Some(c) = self.child_at(x, k)? else { break };
if n.matches_child(self.xml_node(c)?) {
self.xml_node_mut(c)?.parent = None;
} else {
kept.push(c);
self.xml_node_mut(x)?.children.swap(kept, k);
kept += 1;
}
}
self.xml_node_mut(x)?.children = kept;
self.xml_node_mut(x)?.children.truncate(kept);
Ok(())
}
@@ -939,7 +948,7 @@ impl Interp {
self.use_fuel()?;
cur = self.xml_node(a)?.parent;
}
let height = self.node_height(v)?;
let height = self.rooted(&[Value::Object(x), Value::Object(v)], |i| i.node_height(v))?;
if depth.saturating_add(height) > self.options.max_depth {
return Err(Abrupt::Limit(LimitKind::Heap));
}
@@ -971,7 +980,9 @@ impl Interp {
/// The levels of the tree at `root`, counting it.
fn node_height(&mut self, root: ObjId) -> JsResult<usize> {
let mut stack = vec![(root, 0usize)];
let mut held = HeldBytes::new(self);
let mut stack = Vec::new();
push_held(self, &mut held, &mut stack, (root, 0usize))?;
let mut height = 1;
while let Some(&(node, next)) = stack.last() {
self.use_fuel()?;
@@ -980,7 +991,7 @@ impl Interp {
if let Some(top) = stack.last_mut() {
top.1 += 1;
}
stack.push((c, 0));
push_held(self, &mut held, &mut stack, (c, 0))?;
height = height.max(stack.len());
}
None => {
@@ -1049,19 +1060,22 @@ impl Interp {
}
/// Takes the fuel and heap to put every item of `l` under `x` and returns the nodes to place,
/// from [`Self::child_for_item`]. The tree does not change here, so a limit leaves it as it
/// was.
fn reserve_list_insert(&mut self, x: ObjId, l: ObjId) -> JsResult<Vec<ObjId>> {
/// from [`Self::child_for_item`], with the hold on the list of them, which the caller keeps
/// until they are placed. The tree does not change here, so a limit leaves it as it was.
fn reserve_list_insert(&mut self, x: ObjId, l: ObjId) -> JsResult<(Vec<ObjId>, HeldBytes)> {
let count = self.list_len(l)?;
let roots = [Value::Object(x), Value::Object(l)];
let mut held = HeldBytes::new(self);
self.rooted(&roots, |i| held.add(i, size_of::<ObjId>().saturating_mul(count)))?;
let mut items = self.xml_list(l)?.items.clone();
let bytes = XML_REF_BYTES.saturating_mul(items.len());
let roots = [Value::Object(x), Value::Object(l)];
self.rooted(&roots, |i| i.xml_charge(bytes))?;
let len = self.children_len(x)?;
for (j, item) in items.iter_mut().enumerate() {
*item = self.child_for_item(*item, &roots)?;
self.place_fuel(len + j, *item)?;
}
Ok(items)
Ok((items, held))
}
/// Puts `items` at `at` in order, after [`Self::reserve_list_insert`].
@@ -1100,7 +1114,7 @@ impl Interp {
self.insert_node(x, at, *v)?;
}
Content::List(l) => {
let items = self.reserve_list_insert(x, *l)?;
let (items, _held) = self.reserve_list_insert(x, *l)?;
self.place_children(x, at, &items)?;
}
}
@@ -1152,7 +1166,7 @@ impl Interp {
}
Content::List(l) => {
self.check_attach(x, c)?;
let items = self.reserve_list_insert(x, *l)?;
let (items, _held) = self.reserve_list_insert(x, *l)?;
let len = self.children_len(x)?;
self.xml_text_fuel(len)?;
if at < len {
@@ -1229,7 +1243,8 @@ impl Interp {
s.push_units(self, &[SPACE])?;
}
let Some(item) = self.list_at(*l, i)? else { break };
let text = self.node_to_str(item)?.flat();
let text = self.node_to_str(item)?;
let text = flat_charged(self, &text)?;
s.push_units(self, &text)?;
}
let Value::String(s) = s.finish(self) else { return Err(Abrupt::Internal("not a string".into())) };
@@ -1240,9 +1255,9 @@ impl Interp {
let mut found = None;
let count = self.xml_node(x)?.attrs.len();
self.xml_take_fuel(count)?;
let attrs = std::mem::take(&mut self.xml_node_mut(x)?.attrs);
let mut kept = Vec::with_capacity(attrs.len());
for a in attrs {
let mut kept = 0;
for k in 0..count {
let Some(a) = self.xml_node(x)?.attrs.get(k).copied() else { break };
let hit = self.xml_node(a)?.name.as_ref().is_some_and(|q| n.matches(q));
if hit && found.is_some() {
self.xml_node_mut(a)?.parent = None;
@@ -1250,10 +1265,11 @@ impl Interp {
if hit {
found = Some(a);
}
kept.push(a);
self.xml_node_mut(x)?.attrs.swap(kept, k);
kept += 1;
}
}
self.xml_node_mut(x)?.attrs = kept;
self.xml_node_mut(x)?.attrs.truncate(kept);
if let Some(a) = found {
self.xml_node_mut(a)?.value = value;
return Ok(());
@@ -1280,11 +1296,13 @@ impl Interp {
}
let len = self.children_len(x)?;
self.xml_take_fuel(len)?;
self.xml_keep(x);
let mut held = HeldBytes::new(self);
let mut matches = Vec::new();
for k in 0..len {
let Some(child) = self.child_at(x, k)? else { break };
if n.matches_child(self.xml_node(child)?) {
matches.push(child);
push_held(self, &mut held, &mut matches, child)?;
}
}
match c {
@@ -1378,6 +1396,8 @@ impl Interp {
/// The steps of [`Self::resolve_value`], setting `created` to the first element they add.
fn resolve_value_adding(&mut self, x: ObjId, created: &mut Option<ObjId>) -> JsResult<Option<ObjId>> {
self.xml_keep(x);
let mut held = HeldBytes::new(self);
let mut chain = Vec::new();
let mut cur = x;
let base = loop {
@@ -1391,7 +1411,7 @@ impl Interp {
if *name.local == *STAR {
return Ok(None);
}
chain.push(l);
push_held(self, &mut held, &mut chain, l)?;
cur = target;
};
let mut base = base;
@@ -1615,6 +1635,9 @@ impl Interp {
Some(parent) => self.index_in_parent(parent, item)?.map(|q| (parent, q)),
None => None,
};
let count = self.list_len(c)?;
let mut held = HeldBytes::new(self);
self.rooted(&[Value::Object(x), Value::Object(c)], |i| held.add(i, size_of::<ObjId>().saturating_mul(count)))?;
let mut items = self.xml_list(c)?.items.clone();
if place.is_some() {
let roots = [Value::Object(x), Value::Object(c)];
@@ -1682,6 +1705,8 @@ impl Interp {
// ---- copies ----
fn copy_shallow(&mut self, src: ObjId, parent: Option<ObjId>) -> JsResult<ObjId> {
let declared = self.xml_node(src)?.ns.len();
self.xml_charge(XML_NS_BYTES.saturating_mul(declared))?;
let node = self.xml_node(src)?;
let copy = XmlNode {
kind: node.kind,
@@ -1692,11 +1717,11 @@ impl Interp {
children: Vec::new(),
ns: node.ns.clone(),
};
let attrs = node.attrs.clone();
self.xml_charge(XML_NS_BYTES.saturating_mul(copy.ns.len()))?;
let c = self.new_xml_node(copy)?;
self.xml_keep(c);
for a in attrs {
let mut k = 0;
while let Some(a) = self.xml_node(src)?.attrs.get(k).copied() {
k += 1;
self.use_fuel()?;
let n = self.xml_node(a)?;
let attr = XmlNode {
@@ -1721,7 +1746,9 @@ impl Interp {
let top = i.copy_shallow(root, None)?;
i.xml_keep(top);
let mark = i.temp_roots.len();
let mut stack = vec![(root, top, 0usize)];
let mut held = HeldBytes::new(i);
let mut stack = Vec::new();
push_held(i, &mut held, &mut stack, (root, top, 0usize))?;
while let Some(&(src, dst, next)) = stack.last() {
i.use_fuel()?;
i.temp_roots.truncate(mark);
@@ -1733,7 +1760,7 @@ impl Interp {
let c = i.copy_shallow(child, Some(dst))?;
i.xml_charge_link(c)?;
i.xml_node_mut(dst)?.children.push(c);
stack.push((child, c, 0));
push_held(i, &mut held, &mut stack, (child, c, 0))?;
}
None => {
stack.pop();
@@ -1778,9 +1805,18 @@ impl Interp {
// ---- comparison ----
/// The namespace, local name and value of the attribute `id`, flattening a value that is a
/// rope with [`flat_charged`].
fn attr_key(&mut self, id: ObjId) -> JsResult<(Option<Key>, Key, Key)> {
let n = self.xml_node(id)?;
let name = n.name.as_ref().ok_or_else(|| Abrupt::Internal("attribute without a name".into()))?;
let (uri, local, value) = (name.uri.clone(), name.local.clone(), n.value.clone());
Ok((uri, local, flat_charged(self, &value)?))
}
fn same_shallow(&mut self, a: ObjId, b: ObjId) -> JsResult<bool> {
let (x, y) = (self.xml_node(a)?, self.xml_node(b)?);
if x.kind != y.kind || x.attrs.len() != y.attrs.len() || x.children.len() != y.children.len() || x.value.flat() != y.value.flat() {
if x.kind != y.kind || x.attrs.len() != y.attrs.len() || x.children.len() != y.children.len() || x.value.len() != y.value.len() {
return Ok(false);
}
match (&x.name, &y.name) {
@@ -1788,26 +1824,25 @@ impl Interp {
(Some(p), Some(q)) if p.local == q.local && p.uri == q.uri => {}
_ => return Ok(false),
}
if x.attrs.is_empty() {
let (xv, yv) = (x.value.clone(), y.value.clone());
if flat_charged(self, &xv)? != flat_charged(self, &yv)? {
return Ok(false);
}
let count = self.xml_node(a)?.attrs.len();
if count == 0 {
return Ok(true);
}
let count = x.attrs.len();
self.xml_take_fuel(count)?;
let mut held = HeldBytes::new(self);
held.add(self, ATTR_COUNT_BYTES.saturating_mul(count))?;
let (x, y) = (self.xml_node(a)?, self.xml_node(b)?);
let (xa, ya) = (x.attrs.clone(), y.attrs.clone());
let mut counts: HashMap<(Option<Key>, Key, Key), usize> = HashMap::new();
let attr_key = |i: &Self, id: ObjId| -> JsResult<(Option<Key>, Key, Key)> {
let n = i.xml_node(id)?;
let name = n.name.as_ref().ok_or_else(|| Abrupt::Internal("attribute without a name".into()))?;
Ok((name.uri.clone(), name.local.clone(), n.value.flat()))
};
for a in xa {
*counts.entry(attr_key(self, a)?).or_default() += 1;
for k in 0..count {
let Some(id) = self.xml_node(a)?.attrs.get(k).copied() else { break };
*counts.entry(self.attr_key(id)?).or_default() += 1;
}
for b in ya {
match counts.get_mut(&attr_key(self, b)?) {
for k in 0..count {
let Some(id) = self.xml_node(b)?.attrs.get(k).copied() else { break };
match counts.get_mut(&self.attr_key(id)?) {
Some(n) if *n > 0 => *n -= 1,
_ => return Ok(false),
}
@@ -1884,7 +1919,8 @@ impl Interp {
let (xk, yk) = (self.xml_kind(x)?, self.xml_kind(y)?);
let textual = |k| matches!(k, XmlKind::Text | XmlKind::Attribute);
if (textual(xk) && self.node_simple(y)?) || (textual(yk) && self.node_simple(x)?) {
return Ok(Some(self.node_to_str(x)?.flat() == self.node_to_str(y)?.flat()));
let (s, other) = (self.node_to_str(x)?, self.node_to_str(y)?);
return Ok(Some(flat_charged(self, &s)? == flat_charged(self, &other)?));
}
return self.node_equals(x, y).map(Some);
}
@@ -1900,9 +1936,9 @@ impl Interp {
if let Xv::Node(n) = self.classify(x)
&& self.node_simple(n)?
{
let s = self.node_to_str(n)?.flat();
let other = self.to_str(y)?.flat();
return Ok(Some(s == other));
let s = self.node_to_str(n)?;
let other = self.to_str(y)?;
return Ok(Some(flat_charged(self, &s)? == flat_charged(self, &other)?));
}
}
Ok(None)
@@ -1951,7 +1987,9 @@ impl Interp {
/// The nodes under `x` matching `n`, in document order ([[Descendants]], section 9.1.1.8).
pub(crate) fn node_descendants(&mut self, x: ObjId, n: &NameRef, out: ObjId) -> JsResult<()> {
let mut stack = vec![(x, 0usize)];
let mut held = HeldBytes::new(self);
let mut stack = Vec::new();
push_held(self, &mut held, &mut stack, (x, 0usize))?;
if n.attribute {
self.collect_matching(x, n, out)?;
}
@@ -1969,14 +2007,16 @@ impl Interp {
} else if n.matches_child(self.xml_node(child)?) {
self.list_push(out, child)?;
}
stack.push((child, 0));
push_held(self, &mut held, &mut stack, (child, 0))?;
}
Ok(())
}
/// Merges adjacent text and removes empty text under `root` (ECMA-357 section 13.4.4.26).
pub(crate) fn node_normalize(&mut self, root: ObjId) -> JsResult<()> {
let mut stack = vec![(root, 0usize)];
let mut held = HeldBytes::new(self);
let mut stack = Vec::new();
push_held(self, &mut held, &mut stack, (root, 0usize))?;
while let Some(&(x, mut i)) = stack.last() {
self.use_fuel()?;
let Some(child) = self.child_at(x, i)? else {
@@ -2010,7 +2050,7 @@ impl Interp {
frame.1 = i + 1;
}
if kind == XmlKind::Element {
stack.push((child, 0));
push_held(self, &mut held, &mut stack, (child, 0))?;
}
}
Ok(())
@@ -2032,7 +2072,8 @@ impl Interp {
let Some(c) = self.child_at(x, i)? else { break };
let node = self.xml_node(c)?;
if !matches!(node.kind, XmlKind::Comment | XmlKind::Pi) {
let text = node.value.flat();
let text = node.value.clone();
let text = flat_charged(self, &text)?;
out.push_units(self, &text)?;
}
}
@@ -2069,7 +2110,8 @@ impl Interp {
self.use_fuel()?;
let Some(item) = self.list_at(l, i)? else { break };
if !matches!(self.xml_kind(item)?, XmlKind::Comment | XmlKind::Pi) {
let text = self.node_to_str(item)?.flat();
let text = self.node_to_str(item)?;
let text = flat_charged(self, &text)?;
out.push_units(self, &text)?;
}
}
@@ -2107,6 +2149,46 @@ impl Interp {
out.push_units(self, text.get(from..).unwrap_or_default())
}
/// The namespace `stack` declares for `uri`, nearest first, skipping prefixes redeclared nearer.
/// `prefer` picks that prefix when several match; `need_prefix` rejects the default namespace.
/// The set of prefixes it has passed is held against the heap limit.
fn ns_lookup(&mut self, stack: &[XmlNs], uri: &Key, prefer: Option<&Key>, need_prefix: bool) -> JsResult<Option<XmlNs>> {
let mut held = HeldBytes::new(self);
let mut seen: HashSet<&[u16]> = HashSet::new();
let mut best = None;
for ns in stack.iter().rev() {
let Some(p) = &ns.prefix else { continue };
if !insert_held(self, &mut held, &mut seen, &p[..])? {
continue;
}
if ns.uri == *uri && !(need_prefix && p.is_empty()) {
if prefer == Some(p) {
return Ok(Some(ns.clone()));
}
if best.is_none() {
best = Some(ns.clone());
}
}
}
Ok(best)
}
/// The first of `ns0`, `ns1`, ... that no entry of `stack` uses. One of the first
/// `stack.len() + 1` is free, so one pass over `stack` marks the used ones.
fn fresh_prefix(&mut self, stack: &[XmlNs]) -> JsResult<Key> {
self.xml_take_fuel(stack.len() + 1)?;
let mut held = HeldBytes::new(self);
held.add(self, stack.len() + 1)?;
let mut used = vec![false; stack.len() + 1];
for ns in stack {
if let Some(slot) = ns.prefix.as_deref().and_then(prefix_number).and_then(|n| used.get_mut(n)) {
*slot = true;
}
}
let n = used.iter().position(|u| !u).unwrap_or(stack.len());
Ok(key(&format!("ns{n}")))
}
/// Picks the prefix of a name in `scope`, adding a declaration to `scope` when none is there.
fn out_prefix(&mut self, name: &XmlName, scope: &mut Vec<XmlNs>, attribute: bool) -> JsResult<Key> {
self.xml_take_fuel(scope.len() + 1)?;
@@ -2122,13 +2204,13 @@ impl Interp {
if *uri == *key(XML_URI) {
return Ok(key("xml"));
}
if let Some(ns) = ns_lookup(scope, &uri, name.prefix.as_ref(), attribute) {
if let Some(ns) = self.ns_lookup(scope, &uri, name.prefix.as_ref(), attribute)? {
return Ok(ns.prefix.unwrap_or(empty));
}
let prefix = match &name.prefix {
Some(p) if !(attribute && p.is_empty()) && prefix_in_scope(scope, p).is_none() => p.clone(),
_ if !attribute && prefix_in_scope(scope, &empty).is_none() => empty,
_ => fresh_prefix(scope),
_ => self.fresh_prefix(scope)?,
};
self.xml_charge(XML_NS_BYTES)?;
scope.push(XmlNs { prefix: Some(prefix.clone()), uri });
@@ -2143,8 +2225,7 @@ impl Interp {
out.push_units(self, local)
}
/// Writes `x`; an element with children is left open on `stack`.
#[allow(clippy::too_many_arguments)]
/// Writes `x`, and returns the frame of an element with children, which is left open.
fn write_enter(
&mut self,
out: &mut Builder,
@@ -2153,11 +2234,11 @@ impl Interp {
settings: &Settings,
pretty: bool,
scope: &mut Vec<XmlNs>,
stack: &mut Vec<OutFrame>,
) -> JsResult<()> {
) -> JsResult<Option<OutFrame>> {
self.use_fuel()?;
let node = self.xml_node(x)?;
let (kind, value) = (node.kind, node.value.flat());
let (kind, value) = (node.kind, node.value.clone());
let value = flat_charged(self, &value)?;
self.xml_text_fuel(value.len())?;
if pretty {
out.push_repeated(self, &[SPACE], indent)?;
@@ -2165,13 +2246,18 @@ impl Interp {
match kind {
XmlKind::Text => {
let trimmed: &[u16] = if pretty { trim_xml_space(&value) } else { &value };
return self.write_escaped(out, trimmed, false);
self.write_escaped(out, trimmed, false)?;
return Ok(None);
}
XmlKind::Attribute => {
self.write_escaped(out, &value, true)?;
return Ok(None);
}
XmlKind::Attribute => return self.write_escaped(out, &value, true),
XmlKind::Comment => {
out.push_str(self, "<!--")?;
out.push_units(self, &value)?;
return out.push_str(self, "-->");
out.push_str(self, "-->")?;
return Ok(None);
}
XmlKind::Pi => {
let target = self.xml_node(x)?.name.as_ref().map(|n| n.local.clone()).unwrap_or_else(|| Rc::from(Vec::new()));
@@ -2179,47 +2265,46 @@ impl Interp {
out.push_units(self, &target)?;
out.push_units(self, &[SPACE])?;
out.push_units(self, &value)?;
return out.push_str(self, "?>");
out.push_str(self, "?>")?;
return Ok(None);
}
XmlKind::Element => {}
}
let ns_mark = scope.len();
let node = self.xml_node(x)?;
let declared = node.ns.clone();
let attrs = node.attrs.clone();
let declared = node.ns.len();
let name = node.name.clone().ok_or_else(|| Abrupt::Internal("element without a name".into()))?;
self.xml_take_fuel(declared.len().saturating_mul(scope.len() + 1))?;
let mut unnamed = Vec::new();
for ns in declared {
if ns.prefix.is_none() {
unnamed.push(ns);
} else if !scope.iter().any(|a| a.uri == ns.uri && a.prefix == ns.prefix) {
self.xml_charge(XML_NS_BYTES)?;
scope.push(ns);
}
}
for ns in unnamed {
if !scope.iter().any(|a| a.uri == ns.uri) {
let prefix = fresh_prefix(scope);
self.xml_charge(XML_NS_BYTES)?;
scope.push(XmlNs { prefix: Some(prefix), uri: ns.uri });
self.xml_take_fuel(declared.saturating_mul(scope.len() + 1))?;
for unnamed in [false, true] {
let mut k = 0;
while let Some(ns) = self.xml_node(x)?.ns.get(k).cloned() {
k += 1;
if ns.prefix.is_none() != unnamed {
continue;
}
if !unnamed && !scope.iter().any(|a| a.uri == ns.uri && a.prefix == ns.prefix) {
self.xml_charge(XML_NS_BYTES)?;
scope.push(ns);
} else if unnamed && !scope.iter().any(|a| a.uri == ns.uri) {
let prefix = self.fresh_prefix(scope)?;
self.xml_charge(XML_NS_BYTES)?;
scope.push(XmlNs { prefix: Some(prefix), uri: ns.uri });
}
}
}
let prefix = self.out_prefix(&name, scope, false)?;
let mut attr_prefixes = Vec::new();
for &a in &attrs {
self.use_fuel()?;
let an = self.xml_node(a)?.name.clone().ok_or_else(|| Abrupt::Internal("attribute without a name".into()))?;
attr_prefixes.push(self.out_prefix(&an, scope, true)?);
}
out.push_units(self, &[LT])?;
self.write_name(out, &prefix, &name.local)?;
for (&a, attr_prefix) in attrs.iter().zip(&attr_prefixes) {
let an = self.xml_node(a)?;
let local = an.name.as_ref().map(|q| q.local.clone()).unwrap_or_else(|| Rc::from(Vec::new()));
let value = an.value.flat();
let mut k = 0;
while let Some(a) = self.xml_node(x)?.attrs.get(k).copied() {
k += 1;
self.use_fuel()?;
let an = self.xml_node(a)?.name.clone().ok_or_else(|| Abrupt::Internal("attribute without a name".into()))?;
let attr_prefix = self.out_prefix(&an, scope, true)?;
let value = self.xml_node(a)?.value.clone();
let value = flat_charged(self, &value)?;
out.push_units(self, &[SPACE])?;
self.write_name(out, attr_prefix, &local)?;
self.write_name(out, &attr_prefix, &an.local)?;
out.push_units(self, &[EQ, DQUOTE])?;
self.write_escaped(out, &value, true)?;
out.push_units(self, &[DQUOTE])?;
@@ -2239,7 +2324,8 @@ impl Interp {
let count = self.children_len(x)?;
if count == 0 {
scope.truncate(ns_mark);
return out.push_str(self, "/>");
out.push_str(self, "/>")?;
return Ok(None);
}
out.push_units(self, &[GT])?;
let first_is_text = match self.child_at(x, 0)? {
@@ -2248,14 +2334,17 @@ impl Interp {
};
let pretty_children = pretty && (count > 1 || !first_is_text);
let child_indent = if pretty_children { indent.saturating_add(u64::from(settings.indent)) } else { 0 };
stack.push(OutFrame { node: x, next: 0, indent, child_indent, pretty_children, ns_mark, prefix, local: name.local });
Ok(())
Ok(Some(OutFrame { node: x, next: 0, indent, child_indent, pretty_children, ns_mark, prefix, local: name.local }))
}
/// Writes the tree at `root`. Its open elements are held against the heap limit.
fn write_node(&mut self, out: &mut Builder, root: ObjId, indent: u64, settings: &Settings, pretty: bool) -> JsResult<()> {
let mut scope: Vec<XmlNs> = Vec::new();
let mut held = HeldBytes::new(self);
let mut stack: Vec<OutFrame> = Vec::new();
self.write_enter(out, root, indent, settings, pretty, &mut scope, &mut stack)?;
if let Some(frame) = self.write_enter(out, root, indent, settings, pretty, &mut scope)? {
push_held(self, &mut held, &mut stack, frame)?;
}
while let Some(frame) = stack.last_mut() {
let (node, next, child_indent, pretty_children) = (frame.node, frame.next, frame.child_indent, frame.pretty_children);
frame.next += 1;
@@ -2264,7 +2353,9 @@ impl Interp {
if pretty_children {
out.push_units(self, &[LF])?;
}
self.write_enter(out, child, child_indent, settings, pretty && pretty_children, &mut scope, &mut stack)?;
if let Some(frame) = self.write_enter(out, child, child_indent, settings, pretty && pretty_children, &mut scope)? {
push_held(self, &mut held, &mut stack, frame)?;
}
}
None => {
let Some(frame) = stack.pop() else { break };
@@ -2311,8 +2402,11 @@ impl Interp {
self.throw(ErrorType::Syntax, message)
}
fn decode_into(&mut self, src: &[u16], from: usize, to: usize, out: &mut Vec<u16>, attribute: bool) -> JsResult<()> {
/// Decodes `src[from..to]` onto `out`, holding the room it takes in `out` with `held` first. The
/// decoded text is never longer than the source.
fn decode_into(&mut self, src: &[u16], from: usize, to: usize, out: &mut Vec<u16>, attribute: bool, held: &mut HeldBytes) -> JsResult<()> {
self.xml_text_fuel(to.saturating_sub(from))?;
reserve_held(self, held, out, to.saturating_sub(from))?;
let mut pos = from;
while pos < to {
let Some(&c) = src.get(pos) else { break };
@@ -2427,7 +2521,16 @@ impl Interp {
Ok(())
}
fn flush_text(&mut self, text: &mut Vec<u16>, ignore_whitespace: bool, open: &[ObjId], roots: &mut Vec<ObjId>) -> JsResult<()> {
/// Makes a text node of `text`, unless it is empty or ignored white space. Taking the buffer
/// releases `text_held`, its hold.
fn flush_text(
&mut self,
text: &mut Vec<u16>,
text_held: &mut HeldBytes,
ignore_whitespace: bool,
open: &[ObjId],
roots: &mut Vec<ObjId>,
) -> JsResult<()> {
if text.is_empty() {
return Ok(());
}
@@ -2435,16 +2538,22 @@ impl Interp {
text.clear();
return Ok(());
}
let t = self.new_text_node(Str::from_units(std::mem::take(text)), None)?;
let value = Str::from_units(std::mem::take(text));
*text_held = HeldBytes::new(self);
let t = self.new_text_node(value, None)?;
self.place_parsed(t, open, roots)
}
/// Parses the open tag at `src[pos]` and returns the offset after it. An element left open is
/// pushed onto `open` and its name onto `names`, both held with `stack_held`. The attributes,
/// sets and copies it works with are held until it returns.
fn parse_open_tag(
&mut self,
src: &[u16],
pos: usize,
open: &mut Vec<ObjId>,
names: &mut Vec<(usize, usize)>,
stack_held: &mut HeldBytes,
roots: &mut Vec<ObjId>,
) -> JsResult<usize> {
if open.len() >= self.options.max_depth {
@@ -2455,6 +2564,7 @@ impl Interp {
let raw_name = src.get(name_start..name_end).unwrap_or_default();
let (elem_prefix, elem_local) = self.split_name(raw_name)?;
let mut at = name_end;
let mut held = HeldBytes::new(self);
let mut raw_attrs: Vec<(usize, usize, Vec<u16>)> = Vec::new();
let closes = loop {
let before = at;
@@ -2492,27 +2602,31 @@ impl Interp {
return Err(self.syntax("'<' cannot appear in an attribute value"));
}
let mut value = Vec::new();
self.decode_into(src, value_start, value_end, &mut value, true)?;
raw_attrs.push((at, end, value));
self.decode_into(src, value_start, value_end, &mut value, true, &mut held)?;
push_held(self, &mut held, &mut raw_attrs, (at, end, value))?;
at = value_end + 1;
}
}
};
self.xml_charge(XML_NS_BYTES.saturating_mul(raw_attrs.len()))?;
let mut ns: Vec<XmlNs> = Vec::new();
let mut seen_decl: HashSet<Vec<u16>> = HashSet::new();
let xmlns = key("xmlns");
let declares = |raw: &[u16]| raw == &xmlns[..] || raw.strip_prefix(&xmlns[..]).is_some_and(|rest| rest.first() == Some(&COLON));
let decls = raw_attrs.iter().filter(|(start, end, _)| declares(src.get(*start..*end).unwrap_or_default())).count();
let mut ns: Vec<XmlNs> = Vec::with_capacity(decls);
let mut seen_decl: HashSet<&[u16]> = HashSet::new();
for (start, end, value) in &raw_attrs {
let raw = src.get(*start..*end).unwrap_or_default();
if raw == key("xmlns").as_ref() {
if raw == &xmlns[..] {
if !ns_binding_allowed(&[], value) {
return Err(self.syntax(RESERVED_NS_BINDING));
}
if !seen_decl.insert(Vec::new()) {
if !insert_held(self, &mut held, &mut seen_decl, &[] as &[u16])? {
return Err(self.syntax("a namespace is declared twice"));
}
self.xml_charge(string_bytes(0).saturating_add(string_bytes(value.len())))?;
ns.push(XmlNs { prefix: Some(Rc::from(Vec::new())), uri: Rc::from(value.as_slice()) });
} else if let (Some(p), local) = self.split_name(raw)?
&& p == key("xmlns").as_ref()
&& p == &xmlns[..]
{
if value.is_empty() {
return Err(self.syntax("a namespace prefix cannot be declared with an empty URI"));
@@ -2520,9 +2634,10 @@ impl Interp {
if !ns_binding_allowed(local, value) {
return Err(self.syntax(RESERVED_NS_BINDING));
}
if !seen_decl.insert(local.to_vec()) {
if !insert_held(self, &mut held, &mut seen_decl, local)? {
return Err(self.syntax("a namespace is declared twice"));
}
self.xml_charge(string_bytes(local.len()).saturating_add(string_bytes(value.len())))?;
ns.push(XmlNs { prefix: Some(Rc::from(local)), uri: Rc::from(value.as_slice()) });
}
}
@@ -2547,7 +2662,8 @@ impl Interp {
ns,
})?;
self.place_parsed(element, open, roots)?;
let mut unique: HashSet<(Key, Vec<u16>)> = HashSet::new();
let mut unique: HashSet<(Key, &[u16])> = HashSet::new();
let mut own: Option<Vec<XmlNs>> = None;
for (start, end, value) in raw_attrs {
self.use_fuel()?;
let raw = src.get(start..end).unwrap_or_default();
@@ -2557,15 +2673,19 @@ impl Interp {
}
let (auri, stored_prefix): (Key, Key) = match prefix {
Some(p) => {
let own = self.xml_node(element)?.ns.clone();
match self.resolve_prefix(p, &own, open)? {
if own.is_none() {
let count = self.xml_node(element)?.ns.len();
held.add(self, size_of::<XmlNs>().saturating_mul(count))?;
own = Some(self.xml_node(element)?.ns.clone());
}
match self.resolve_prefix(p, own.as_deref().unwrap_or_default(), open)? {
Some(u) => (u, Rc::from(p)),
None => return Err(self.syntax("an XML name uses an undeclared namespace prefix")),
}
}
None => (Rc::from(Vec::new()), Rc::from(Vec::new())),
};
if !unique.insert((auri.clone(), local.to_vec())) {
if !insert_held(self, &mut held, &mut unique, (auri.clone(), local))? {
return Err(self.type_error("an XML element has two attributes with the same name"));
}
let a = self.new_xml_node(XmlNode {
@@ -2581,8 +2701,8 @@ impl Interp {
self.xml_node_mut(element)?.attrs.push(a);
}
if !closes {
open.push(element);
names.push((name_start, name_end));
push_held(self, stack_held, open, element)?;
push_held(self, stack_held, names, (name_start, name_end))?;
}
Ok(at)
}
@@ -2605,10 +2725,14 @@ impl Interp {
}
/// Parses XML text into nodes, kept reachable through the temporary roots until the caller's
/// scope ends.
/// scope ends. The open elements and their names, the text being gathered and each tag's
/// attributes are held as they grow. The allowance held from the source's length covers the
/// copy a value's units take as it becomes a string, before the node that holds it is charged.
pub(crate) fn xml_parse(&mut self, source: &[u16]) -> JsResult<Vec<ObjId>> {
let mut held = HeldBytes::new(self);
held.add(self, string_bytes(source.len()).saturating_mul(2))?;
let mut stack_held = HeldBytes::new(self);
let mut text_held = HeldBytes::new(self);
let settings = self.xml_settings()?;
let mut src = source;
let offset_start = skip_xml_space(src, 0);
@@ -2638,7 +2762,7 @@ impl Interp {
self.xml_text_fuel(end - pos)?;
return Err(self.syntax("']]>' cannot appear in XML text"));
}
self.decode_into(src, pos, end, &mut text, false)?;
self.decode_into(src, pos, end, &mut text, false, &mut text_held)?;
pos = end;
continue;
}
@@ -2650,11 +2774,13 @@ impl Interp {
};
self.xml_text_fuel(end - pos)?;
self.check_units(src, pos + 9, end)?;
text.extend_from_slice(src.get(pos + 9..end).unwrap_or_default());
let section = src.get(pos + 9..end).unwrap_or_default();
reserve_held(self, &mut text_held, &mut text, section.len())?;
text.extend_from_slice(section);
pos = end + 3;
continue;
}
self.flush_text(&mut text, settings.ignore_whitespace, &open, &mut roots)?;
self.flush_text(&mut text, &mut text_held, settings.ignore_whitespace, &open, &mut roots)?;
if rest.starts_with(&parser_units("<!--")) {
let Some(end) = Self::find_from(src, pos + 4, &parser_units("--")) else {
self.xml_text_fuel(src.len() - pos)?;
@@ -2666,7 +2792,7 @@ impl Interp {
}
self.check_units(src, pos + 4, end)?;
if !settings.ignore_comments {
let value = Str::from_units(src.get(pos + 4..end).unwrap_or_default().to_vec());
let value = Str::from_rc(Rc::from(src.get(pos + 4..end).unwrap_or_default()));
let n = self.new_xml_node(XmlNode {
kind: XmlKind::Comment,
name: None,
@@ -2698,7 +2824,7 @@ impl Interp {
while from < end && src.get(from).copied().is_some_and(is_xml_space) {
from += 1;
}
let value = Str::from_units(src.get(from..end).unwrap_or_default().to_vec());
let value = Str::from_rc(Rc::from(src.get(from..end).unwrap_or_default()));
let name = XmlName { uri: Some(Rc::from(Vec::new())), local: Rc::from(target), prefix: None };
let n = self.new_xml_node(XmlNode {
kind: XmlKind::Pi,
@@ -2715,13 +2841,13 @@ impl Interp {
} else if rest.get(1) == Some(&SLASH) {
pos = self.parse_close_tag(src, pos, &mut open, &mut names)?;
} else {
pos = self.parse_open_tag(src, pos, &mut open, &mut names, &mut roots)?;
pos = self.parse_open_tag(src, pos, &mut open, &mut names, &mut stack_held, &mut roots)?;
}
}
if !open.is_empty() {
return Err(self.syntax("an XML element is not closed"));
}
self.flush_text(&mut text, settings.ignore_whitespace, &open, &mut roots)?;
self.flush_text(&mut text, &mut text_held, settings.ignore_whitespace, &open, &mut roots)?;
Ok(roots)
}
@@ -2752,7 +2878,8 @@ impl Interp {
(name.uri.clone(), name.local.clone())
}
v => {
let local = self.to_str(v)?.flat();
let local = self.to_str(v)?;
let local = flat_charged(self, &local)?;
let uri = if *local == *STAR {
None
} else if attribute {
@@ -2770,9 +2897,15 @@ impl Interp {
Value::Object(o) => match self.heap.object(*o).map(|ob| &ob.class) {
Some(Class::Namespace(ns)) => Some(ns.uri.clone()),
Some(Class::QName { name, .. }) => name.uri.clone(),
_ => Some(self.to_str(&q)?.flat()),
_ => {
let s = self.to_str(&q)?;
Some(flat_charged(self, &s)?)
}
},
_ => Some(self.to_str(&q)?.flat()),
_ => {
let s = self.to_str(&q)?;
Some(flat_charged(self, &s)?)
}
};
}
let id = self.new_reference_name(XmlName { uri, local, prefix: None }, attribute)?;
@@ -2816,7 +2949,8 @@ impl Interp {
let result = match kind {
0 if self.is_xml(&v) => self.xml_string_of(&v, Some(false))?,
0 | 1 => {
let text = self.to_str(&v)?.flat();
let text = self.to_str(&v)?;
let text = flat_charged(self, &text)?;
let mut out = Builder::new();
self.write_escaped(&mut out, &text, kind == 1)?;
let Value::String(s) = out.finish(self) else { return Err(Abrupt::Internal("not a string".into())) };
@@ -2991,7 +3125,8 @@ impl Interp {
_ => {}
}
}
Ok(self.to_str(v)?.flat())
let s = self.to_str(v)?;
flat_charged(self, &s)
}
pub(crate) fn op_set_default_namespace(&mut self) -> JsResult<()> {
crates/js/src/xml_tests.rs
+245
View File
@@ -1185,6 +1185,251 @@ fn a_limit_during_normalize_does_not_repeat_text() {
assert!(fuel_stops(setup, "x.normalize()", check) > 0);
}
/// A realm whose trees may be 100,000 levels deep, with `squeeze`, after `setup` ran with no limit.
fn deep_realm(setup: &str) -> Realm {
let mut r = Realm::with_options(Options { max_depth: 100_000, ..e4x() }, Limits::default());
let interp = r.interp();
let global = interp.global;
assert!(crate::builtins::method(interp, global, "squeeze", squeeze, 1).is_ok());
r.run(setup).unwrap();
r
}
/// How `action` ends after `setup` in a [`deep_realm`], asserting that `check` is true after it
/// with the limits lifted.
fn squeezed(setup: &str, action: &str, check: &str) -> Result<JsValue, RunErrorKind> {
let mut r = deep_realm(setup);
let got = r.run(action).map_err(|e| e.kind);
r.set_limits(Limits::default());
assert_eq!(r.run(check), Ok(JsValue::Bool(true)), "{action}: {got:?}");
got
}
/// The least spare heap, found by bisection, at which `action` ends without the heap limit after
/// `setup` in a [`deep_realm`].
fn least_room(setup: &str, action: &str) -> usize {
let ends = |spare: usize| match deep_realm(setup).run(&format!("squeeze({spare}); {action}")) {
Ok(_) => true,
Err(e) if e.kind == RunErrorKind::Limit(LimitKind::Heap) => false,
Err(e) => panic!("{action}: {e:?}"),
};
let (mut lo, mut hi) = (0usize, 1usize << 22);
assert!(ends(hi), "{action}");
while lo < hi {
let mid = lo + (hi - lo) / 2;
if ends(mid) {
hi = mid;
} else {
lo = mid + 1;
}
}
lo
}
/// Scripts that make `x`, an `r` element over 2,000 elements named `d`, nested in the first and
/// side by side, named `wide_name`, in the second. Both also make `t`, an empty element, and turn
/// pretty printing off.
fn deep_and_wide(wide_name: &str) -> (String, String) {
let common = "XML.prettyPrinting = false; var t = <t/>; var s = '';";
let deep = format!(
"{common} for (var i = 0; i < 2000; i++) s += '<d>'; for (var i = 0; i < 2000; i++) s += '</d>'; var x = new XML('<r>' + s + '</r>');"
);
let wide = format!("{common} for (var i = 0; i < 2000; i++) s += '<{wide_name}/>'; var x = new XML('<r>' + s + '</r>');");
(deep, wide)
}
#[test]
fn tree_walks_hold_their_stacks_against_the_heap_limit() {
// A walk keeps a frame for each open level, so the deep tree needs room for 2,000 frames that
// the wide tree, with as many nodes and the same markup length, does not.
let mut wrong = Vec::new();
for (action, wide_name, frame) in [
("x.copy(); 1", "d", 8),
("x..z.length()", "d", 8),
("x.normalize(); 1", "d", 8),
("t.appendChild(x); 1", "d", 8),
("x.toXMLString().length", "dddd", 32),
] {
let (deep, wide) = deep_and_wide(wide_name);
let (d, w) = (least_room(&deep, action), least_room(&wide, action));
if d < w + 2000 * frame {
wrong.push(format!("{action}: deep {d}, wide {w}"));
}
}
assert!(wrong.is_empty(), "{wrong:#?}");
let (deep, _) = deep_and_wide("d");
let same = "var y = x.copy(); var a = x, b = y, n = 0; while (a.d.length()) { a = a.d[0]; b = b.d[0]; n++; if (b.parent().d[0] !== b) break; } \
n === 2000 && y.parent() === null && y == x && y.toXMLString() === x.toXMLString()";
assert_eq!(squeezed(&deep, same, "true"), Ok(JsValue::Bool(true)));
}
#[test]
fn assignments_hold_the_matches_and_targets_they_collect() {
let heap = Err(RunErrorKind::Limit(LimitKind::Heap));
let wide = "var x = <r/>; for (var i = 0; i < 4000; i++) x.appendChild(<b/>);";
let kept = "x.*.length() === 4000 && x.b[3999].parent() === x";
assert_eq!(squeezed(wide, "squeeze(4096); x.b = 'v'; 1", kept), heap);
let first = "var old = x.b[0]; x.b = 'v'; x.*.length() === 1 && x.b[0] === old && x.b == 'v'";
assert_eq!(squeezed(wide, first, "x.b[0].parent() === x"), Ok(JsValue::Bool(true)));
// `l` is an empty list 4,000 `.b` steps below `x.a`, which holds two elements, so the
// assignment adds nothing.
let chain = "var x = <r><a/><a/></r>; var l = x.a; for (var i = 0; i < 4000; i++) l = l.b;";
let same = "x.*.length() === 2 && x.a.*.length() === 0";
assert_eq!(squeezed(chain, "squeeze(4096); l.c = 'v'; 1", same), heap);
assert_eq!(squeezed(chain, "l.c = 'v'; 1", same), Ok(JsValue::Number(1.0)));
}
#[test]
fn serializing_charges_the_text_it_flattens() {
// `k` is a rope of 32,768 units. Both trees hold the same 65,536 units of text, as a rope in the
// first and as flat text in the second.
let k = "var k = 'x'; for (var i = 0; i < 15; i++) k += k; XML.prettyPrinting = false;";
let rope = format!("{k} var x = <r/>; x.appendChild(k); x.appendChild(k); x.normalize();");
let flat = format!("{k} var x = new XML('<r>' + k + k + '</r>');");
let action = "x.toXMLString().length";
let (r, f) = (least_room(&rope, action), least_room(&flat, action));
assert!(r >= f + 100_000, "rope {r}, flat {f}");
let attr = format!("{k} var x = <r/>; x.@a = k + k;");
let same = "var s = x.toXMLString(); s.length > 65536 && s.indexOf(k + k) > 0 && new XML(s).toXMLString() === s";
for setup in [rope, attr] {
assert_eq!(squeezed(&setup, same, "true"), Ok(JsValue::Bool(true)), "{setup}");
}
}
/// A script that makes `x`, a `r:r` element over 2,000 nested `r:d` elements in the same
/// namespace, each of which declares one more prefix: a new one each time, or `q10000` each time.
fn nested_prefixes(same: bool) -> String {
let prefix = if same { "10000" } else { "' + (10000 + i) + '" };
format!(
"XML.prettyPrinting = false; var s = ''; for (var i = 0; i < 2000; i++) s += '<r:d xmlns:q{prefix}=\"u' + (10000 + i) + '\">'; \
for (var i = 0; i < 2000; i++) s += '</r:d>'; var x = new XML('<r:r xmlns:r=\"R\">' + s + '</r:r>');"
)
}
#[test]
fn serializing_holds_the_prefixes_a_lookup_passes() {
// Finding the `r` prefix from the deepest element passes 2,000 prefixes when each level
// declares a new one, and one when they all declare `q10000`.
let action = "x.toXMLString().length";
let (new, same) = (least_room(&nested_prefixes(false), action), least_room(&nested_prefixes(true), action));
assert!(new >= same + 2000 * 16, "new {new}, same {same}");
let check = "var s = x.toXMLString(); s.indexOf('<r:r xmlns:r=\"R\"><r:d xmlns:q10000=\"u10000\"><r:d xmlns:q10001=\"u10001\">') === 0 && new XML(s).toXMLString() === s";
assert_eq!(squeezed(&nested_prefixes(false), check, "true"), Ok(JsValue::Bool(true)));
}
#[test]
fn a_new_prefix_is_the_first_free_ns_number() {
let setup = "XML.prettyPrinting = false; var n = new Namespace('urn:z');";
for (decls, prefix) in [
("", "ns0"),
(" xmlns:ns0=\"a\" xmlns:ns1=\"b\" xmlns:ns3=\"c\" xmlns:ns02=\"d\"", "ns2"),
(" xmlns:ns0=\"a\" xmlns:ns1=\"b\" xmlns:nsx=\"c\" xmlns:ns2=\"d\"", "ns3"),
] {
truth(&format!(
"{setup} var x = <r{decls}/>; x.@n::a = 'v'; var s = x.toXMLString(); \
s.indexOf(' {prefix}:a=\"v\"') > 0 && s.indexOf(' xmlns:{prefix}=\"urn:z\"') > 0 && new XML(s).@n::a == 'v'"
));
}
truth(&format!(
"{setup} var s = ''; for (var i = 0; i < 500; i++) s += ' xmlns:ns' + i + '=\"u' + i + '\"'; var x = new XML('<r' + s + '/>'); x.@n::a = 'v'; \
x.toXMLString().indexOf(' ns500:a=\"v\"') > 0"
));
}
/// A script that makes `x`, an `r` element over 4,000 `b` elements, `l`, the list of them, and
/// `t`, an empty element.
const MANY_B: &str = "XML.prettyPrinting = false; var x = <r/>; for (var i = 0; i < 4000; i++) x.appendChild(<b/>); var l = x.b; var t = <t/>;";
#[test]
fn inserting_a_list_holds_its_copy_of_the_items() {
// Both actions move the same 4,000 elements under `t`. The first charges the links, 8 bytes
// each, before any element leaves `x`, and holds its copy of the list's items, 4 bytes each;
// the second moves them one at a time.
let list = least_room(MANY_B, "t.prependChild(l); 1");
let one_by_one = least_room(MANY_B, "for (var i = 0; i < 4000; i++) t.prependChild(l[i]); 1");
assert!(list >= one_by_one + 4000 * (8 + 2), "list {list}, one by one {one_by_one}");
let moved = "t.prependChild(l); t.*.length() === 4000 && x.*.length() === 0 && t.b[3999].parent() === t";
assert_eq!(squeezed(MANY_B, moved, "true"), Ok(JsValue::Bool(true)));
}
#[test]
fn assigning_a_list_to_a_list_item_holds_its_copy_of_the_items() {
// Both actions put the 4,000 elements of `l` under `m`. The first, through the list `m.a`,
// also makes a list of them, 8 bytes each, and holds a copy of its items, 4 bytes each.
let setup = format!("{MANY_B} var m = <m><a/></m>; var la = m.a;");
let through_list = least_room(&setup, "la[0] = l; 1");
let direct = least_room(&setup, "m.prependChild(l); 1");
assert!(through_list >= direct + 4000 * 10, "through the list {through_list}, direct {direct}");
let placed = "la[0] = l; m.*.length() === 4000 && la.length() === 4000 && la[3999] === m.b[3999] && x.*.length() === 0";
assert_eq!(squeezed(&setup, placed, "true"), Ok(JsValue::Bool(true)));
}
#[test]
fn parsing_holds_its_open_elements_and_attribute_scratch() {
// The deep markup leaves 2,000 elements open at once and the wide markup, of the same
// length, leaves none. The open elements and their names take 2,048 times 20 bytes, and the
// longer names of the wide markup take 2,000 times 6 bytes more.
let (deep, wide) = deep_and_wide("dddd");
let action = "new XML('<r>' + s + '</r>'); 1";
let (d, w) = (least_room(&deep, action), least_room(&wide, action));
// Both elements declare 2,001 namespaces and have one attribute. Resolving the prefix of the
// attribute in the first reads a copy of the element's namespaces.
let decls = "var d = ''; for (var i = 10000; i < 12000; i++) d += ' xmlns:n' + i + '=\"u\"';";
let prefixed = least_room(decls, "new XML('<r' + d + ' xmlns:p=\"P\" p:a=\"v\"/>'); 1");
let plain = least_room(decls, "new XML('<r' + d + ' xmlns:p=\"P\" pa=\"v\"/>'); 1");
assert!(d >= w + 2000 * 8 && prefixed >= plain + 2000 * 32, "deep {d}, wide {w}, prefixed {prefixed}, plain {plain}");
truth(&format!(
"{decls} var x = new XML('<r' + d + ' xmlns:p=\"P\" p:a=\"v\" b=\"w\"/>'); x.@b == 'w' && x.namespaceDeclarations().length == 2001"
));
}
#[test]
fn flattening_a_rope_in_xml_code_is_charged() {
let k = "var k = 'x'; for (var i = 0; i < 15; i++) k += k; XML.prettyPrinting = false; var x = <r/>, y = <r/>;";
// Each pair holds the same 65,536 units as ropes in the first script and as flat text in the
// second: the text of `x` and `y`, an attribute of `x` and `y`, and the string `s2`.
let text = (
format!("{k} for each (var e in [x, y]) {{ e.appendChild(k); e.appendChild(k); e.normalize(); }}"),
format!("{k} x = new XML('<r>' + k + k + '</r>'); y = new XML('<r>' + k + k + '</r>');"),
);
let attr = (format!("{k} x.@a = k + k; y.@a = k + k;"), format!("{k} x = new XML('<r a=\"' + k + k + '\"/>'); y = new XML(x.toXMLString());"));
let o = "var o = { toString: function () { return s2; } };";
let string = (format!("{k} var s2 = k + k; {o}"), format!("{k} var s2 = new XML('<r>' + k + k + '</r>').toString(); {o}"));
let mut wrong = Vec::new();
for ((rope, flat), action) in [
(&text, "x == y"),
(&attr, "x == y"),
(&text, "x.text()[0] == y.text()[0]"),
(&text, "x.text()[0] == 'z'"),
(&string, "<e/> == s2"),
(&text, "x.toString().length"),
(&text, "x.text().toString().length"),
(&string, "x.@[s2].length()"),
(&string, "x.s2::b.length()"),
(&string, "x.o::b.length()"),
(&string, "default xml namespace = s2; 1"),
(&string, "x[s2].length()"),
(&string, "x[o].length()"),
] {
let (r, f) = (least_room(rope, action), least_room(flat, action));
if r < f + 100_000 {
wrong.push(format!("{action}: rope {r}, flat {f}"));
}
}
// A name read from a flat string is a copy of its units after `@` for an attribute, and shares
// the string for an element.
let names = format!("{} var at = new XML('<r>@' + k + k + '</r>').toString();", string.1);
let (a, e) = (least_room(&names, "x[at].length()"), least_room(&names, "x[s2].length()"));
if a < e + 100_000 {
wrong.push(format!("names: attribute {a}, element {e}"));
}
assert!(wrong.is_empty(), "{wrong:#?}");
let same = "x == y && x.text()[0] == y.text()[0] && x.toString() === s && x.text().toString() === s && !(x.text()[0] == 'z')";
for setup in [&text.0, &text.1] {
assert_eq!(squeezed(&format!("{setup} var s = k + k;"), same, "true"), Ok(JsValue::Bool(true)), "{setup}");
}
}
/// Runs `action` as [`heap_stops`] does, and asserts that every run completes or stops at the heap
/// limit and that `check` is true after it. Returns how many runs stopped.
fn heap_stops_only(setup: &str, action: &str, check: &str) -> usize {
docs/target-app-parity.md
+1
View File
@@ -811,6 +811,7 @@ Moved from the ROADMAP's "Shipped so far" on 2026-10-10. Update it when a featur
- `vectorcraft-js` has `String`, `Number`, `RegExp` and `Date` as ECMAScript 5.1 defines them (sections 15.5, 15.7, 15.9 and 15.10, https://262.ecma-international.org/5.1/), with Annex B's `substr`, `getYear`, `setYear` and `toGMTString`. `toFixed`, `toExponential` and `toPrecision` format a number from its exact decimal value, and `toString(radix)` writes any base from 2 to 36. Regular expressions (flags `g`, `i` and `m`) run on a backtracking matcher whose choice points and capture state count against the heap limit and which takes fuel for every step, so a catastrophic pattern stops at the fuel limit and 50,000 nested groups give a `SyntaxError` at the default depth limit. A string method that would build a string past the heap limit stops before it allocates. `Date` reads the clock and time zone the host sets (`Realm::set_clock`, `Realm::set_time_zone`; local time is UTC by default). `cargo xtask test262` runs the suite's `built-ins` tests for the four (https://github.com/tc39/test262): of the runs of the ES5 tests that need only the built-ins present, 1,379 of 1,411 pass for `String`, 298 of 298 for `Number`, 954 of 988 for `RegExp` and 8 of 8 for `Date`. The others expect later editions' rules (built-in functions whose `length` can be deleted, flags given with a RegExp object, a negative `lastIndex`, `RegExp.prototype` as a plain object, and `RegExp.prototype`'s `global`, `ignoreCase` and `multiline` properties that can be deleted). Tests: `crates/js/src/text_tests.rs`, `crates/js/tests/test262.rs`.
- With `Options::e4x` set, an E4X name standing alone (`@a`, `@ns::a`, `@[expr]`, `ns::a`, `*`) inside `with` or a filter predicate can be assigned, with `=` or a compound operator, incremented, decremented, deleted and used as a `for` or `for each` variable, as well as read. Each of these uses the nearest XML or XMLList value in scope that holds the name and evaluates the name once. When no XML value in scope holds the name, the read or write throws a `ReferenceError` before the right-hand side runs and creates no child or attribute, as ECMA-357 section 11.1 (https://ecma-international.org/publications-and-standards/standards/ecma-357/) specifies, so a filter such as `list.(@id == '1')` throws when an item has no `@id` and no outer XML value holds one. Tests: `crates/js/src/xml_tests.rs`.
- `vectorcraft-js` runs ECMAScript for XML (ECMA-357, 2nd edition, https://ecma-international.org/publications-and-standards/standards/ecma-357/) when `Options::e4x` is set, which it is not by default: XML and XMLList literals with embedded `{}` expressions, the `..`, `.@`, `.*`, `.()` and `::` selectors, assignment and `delete` on selected nodes, `for each`, `default xml namespace`, and the `XML`, `XMLList`, `Namespace` and `QName` objects with their methods and settings. `xml.xpath(expression[, variables])` evaluates an XPath 1.0 expression (https://www.w3.org/TR/1999/REC-xpath-19991116/) with all 13 axes and the 27 core functions, and returns the selected nodes as an XMLList in document order without repeats. Each property of `variables` binds the variable of its name, an XMLList runs the expression on each of its elements in turn, and a prefix in a name test resolves through the context element's in-scope namespaces. An expression that breaks the grammar gives a `SyntaxError`; an unknown function, a wrong argument count, an unbound prefix or variable, and a result that is not a node-set or that holds the root or a namespace node give a `TypeError`. `id()` selects nothing, and evaluation uses no native stack, takes fuel and counts against the heap limit. `<!DOCTYPE>` in an XML string gives a `SyntaxError`. An XML declaration at the start of an XML string, after optional white space, is checked against XML 1.0's grammar (https://www.w3.org/TR/xml/#NT-XMLDecl): a version of `1.` and digits, then an optional encoding name and an optional standalone `yes` or `no`, in that order; any other declaration, or one with a character XML does not allow, gives a `SyntaxError`. The parser takes fuel for the length of every name, white space run and declaration it scans, and for the text it searched when a quote, CDATA section, comment or processing instruction is not closed. Parsing, serializing, comparing and copying trees use no native stack, and each takes fuel and counts against the heap limit. Inserting a node into itself or its descendants throws a `TypeError` before the tree changes, a literal or XML string nested past `Options::max_depth` gives a `SyntaxError`, and a tree built past that depth stops at the heap limit. Tests: `crates/js/src/xml_tests.rs`.
- Scripts: assigning to a named child of an XML element (`x.b = v`) counts its list of the matching children against the heap limit, and assigning through an empty XMLList counts the chain of targets it climbs, so an assignment that would not fit stops before the tree changes. `copy()`, `toXMLString()`, `normalize()`, the `..` selector and inserting a node count the stack they keep for each open level of the tree, `copy()` and `toXMLString()` read an element's attributes in place instead of copying the list, `toXMLString()` also reads its namespace declarations in place, and `copy()` counts the declarations it copies before it copies them. `toXMLString()` counts the flat copy it makes of text or an attribute value built by joining strings. Deleting children or attributes by name, and assigning an attribute, remove the nodes in place instead of building a second list. `toXMLString()` counts the set of prefixes it passes looking up a namespace, and picks a new `ns` prefix in one pass over the declarations in scope. Inserting an XMLList, and assigning one to an item of an XMLList, count their copy of the list's items. Parsing counts the open elements, the text it gathers, each tag's attributes and the sets that check them, and the namespace declarations it makes. Comparing XML values, converting them to strings, and reading a name, a namespace or embedded text from a string built by joining strings count the flat copy they make; an attribute name read from a string counts its copy. Tests: `crates/js/src/xml_tests.rs`.
- **UI:** Illustrator 2026 layout restyled to measured values:
- Medium Dark theme, categorized and Advanced toolbars (a tool group's flyout opens on a right press, a long press or a press on its corner triangle), 35 pt document tabs, 33 pt panel tabs.
- Hint bar, contextual task bar, 19 dock panels with ≡ menus.