Always use what the old implementation called the "IO" allocation
scheme, allowing to move the content pointer past the initial
allocation. This is inexpensive and allows efficient shrinking.
Optimize xmlBufGrow, reusing shrunken memory as much as possible.
Simplify xmlBufAdd.
Make xmlBufBackToBuffer return an error on overflow.
Make "size" exclude the terminating NULL byte.
Always provide an initial size.
Reintroduce static buffers.
Remove xmlBufResize and several other functions.
Revert a change from d025cfbb and don't overwrite ID table entries, so
that the first attribute will be returned if there are duplicate IDs.
This requires two other changes:
- Attributes in entity content are never added to the ID table. This
seems reasonable.
- Remove the optimization to skip ID lookup when copying and the target
document has an empty ID table. This also seems more correct since the
document could have ID declarations nevertheless or we could be
copying xml:ids into the document for the first time.
Fixes#757.
This option would allow for a smaller, but mostly useless minimal build.
But it complicates the symbol availability logic in an insane way and
requires specialized tools like our custom C parser in doc/apibuild.py.
See #717.
It's possible to create references to predefined entities using the tree
API. This edge case was exposed by making predefined entities const in
commit 63ce5f9a.
xmlNewCharRef also tries to handle strings like '&name;' but in
xmlNodeParseContentInternal, we really want to use the possibly invalid
name without modification. Otherwise, content like '&"' could
create a reference to a predefined entity.
Some users set href to NULL to unset a namespace without deleting it.
Also change the duplicate check in xmlNewNs which must agree with
xmlSearchNs.
Short-lived regression from f960c60d.
Commit 9e1c72da from 2001 introduced a bug where xmlAddPrevSibling and
xmlAddNextSibling would only try to merge text nodes with one of its
new siblings. Commit 4ccd3eb8 fixed this bug but unfortunately, lxml
and possibly other downstream code depend on text nodes not being
merged.
To avoid breaking downstream code while still having somewhat
consistent API behavior, it's probably best to make these functions
never coalesce text nodes.
Make xmlAddChild unlink the child before insertion. Originally, linked
children would most likely cause tree corruption. The first fix
disallowed linked nodes, but there are cases where insertion of such
nodes could succeed.
Don't abort if the node is already a child of parent. In this case,
the node will be moved to the end of the child list.
xmlUnlinkNode also removes references to DTD nodes which shouldn't be
done when moving nodes within a document. Introduce a new function
xmlUnlinkNodeInternal which only unlinks a node from the tree.
Remove references to DTD nodes in xmlNodeSetDoc. Note that moving
element and attribute declarations to another document will still leave
references in the source document.
After the failed experiment with a static XML namespace, introduce
versions of xmlSearchNs that report malloc failures.
Optimize the no-document case by only adding the XML namespace
declaration if it wasn't found in an ancestor.
Replace xmlStringGetNodeList and xmlStringLenGetNodeList with
xmlNodeParseContentInternal which also updates an optional parent
node.
Don't look up entities a second time via xmlNewReference.
Factor out xmlBufGetEntityRefContent and xmlBufGetChildContent.
Also allow entity declarations.
Optimize single text children.
Ignore missing or recursive entities silently.
Prefer xmlNodeGetContent over xmlNodeListGetString.
Check for entity cycles in xmlBufGetNodeContent.
Use children pointer of entity reference nodes if available to look up
entities.
