* lisp/org-element.el: Add commentary explaining some regexp-related
optimizations useful for the parser.
Link: https://debbugs.gnu.org/cgi/bugreport.cgi?bug=63225
(org-element--headline-deferred):
(org-element-item-parser):
(org-element-item-interpreter):
(org-element--list-struct):
(org-element-plain-list-parser):
(org-element-example-block-parser):
(org-element-fixed-width-interpreter):
(org-element-paragraph-parser):
(org-element-src-block-parser):
(org-element-table-parser):
(org-element--parse-generic-emphasis):
(org-element-export-snippet-interpreter):
(org-element-link-parser):
(org-element--current-element):
(org-element--collect-affiliated-keywords):
(org-element-parse-buffer):
(org-element-normalize-string):
(org-element-normalize-contents):
(org-element--parse-to):
(org-element--cache-before-change):
(org-element--cache-for-removal):
(org-element-context): Avoid storing match data unless strictly
necessary. Explain the necessity in places where we have to use
`save-match-data'. Prefer `looking-at-p' (does not alter match data)
to `looking-at'. Simplify regexps. Update docstrings clearly
indicating when match data might be modified.
* lisp/org.el:
(org-offer-links-in-entry):
* lisp/ob-exp.el (org-babel-exp-process-buffer):
* lisp/org-agenda.el: Fix places where we need to protect match data.
* lisp/org-element.el (org-element--headline-parent-deferred): New
function to calculate :parent of a heading on demand.
(org-element--parse-to): Assign deferred :parent when cache is
inactive.
* lisp/org-element.el (org-element-begin):
(org-element-end):
(org-element-contents-begin):
(org-element-contents-end):
(org-element-post-affiliated):
(org-element-post-blank): New functions for fast access to frequently
used element properties. The functions are `setf'-able.
Update all the relevant code in Org to use the new functions instead
of genetic property queries.
* lisp/org-element.el (org-element-parse-buffer): Do not force-undefer
when new optional argument is provided.
* lisp/org-lint.el (org-lint--generate-reports):
* lisp/org.el (org-unindent-buffer):
* lisp/ox.el (org-export--missing-definitions):
(org-export--annotate-info): Do not undefer when unnecessary.
Adjust all the declare forms.
* lisp/org-element-ast.el (org-element-ast-map):
* lisp/org-element.el (org-element-map): Treat TYPES t as all possible
types. Add new optional parameter to no resolve deferred while
traversing the AST.
* lisp/org-element.el (org-element--substring):
(org-element--unescape-substring): New functions to be used for
deferred buffer substring calculation.
(org-element--headline-raw-value):
(org-element-comment-block-parser):
(org-element-example-block-parser):
(org-element-export-block-parser):
(org-element-latex-environment-parser):
(org-element-src-block-parser):
(org-element-table-parser):
(org-element--parse-generic-emphasis):
(org-element-inline-babel-call-parser):
(org-element-latex-fragment-parser):
(org-element-export-snippet-parser): Defer large buffer substrings.
* lisp/org-element.el (org-element--headline-raw-value): New function
calculating :raw-value property dynamically.
(org-element-headline-parser): Use the new function, deferring
:raw-value to reduce memory footprint of the cache.
* lisp/org-element.el (org-element-map): Use `org-element-ast-map' to
traverse the AST. Update the docstring adding new features from
`org-element-ast-map' - FUN can now throw `:org-element-skip' signal
to skip current node and its child nodes. Clarify the structure of
INFO argument. Reverse the traverse order for dual keywords -
secondary value is now traversed after the main value.
* testing/lisp/test-org-element.el (test-org-element/map): Update the test
reflecting the change in dual keyword traversal order.
* lisp/org-element.el (org-element-parse-buffer): Resolve all the
deferred values in the string. If not, we might leave pointers to
killed buffer.
(org-element--parse-elements): Resolve deferred in objects.
(org-element--object-lex): Store :buffer property.
* lisp/org-macro.el (org-macro--find-date): Do not try to print
:buffer property.
* lisp/org-element.el (org-element--cache-persist-before-write):
(org-element--cache-persist-after-read): Clear and restore
non-printable buffer objects in :buffer property.
* lisp/org-element.el: Do not reverse affiliated keyword order.
(org-element-map): Remove special arrangement to restore keyword
ordering.
(org-element--interpret-affiliated-keywords): Obey the new keyword order.
* lisp/ox.el (org-export-get-caption):
* testing/lisp/test-org-element.el (test-org-element/affiliated-keywords-parser):
* testing/lisp/test-org-element.el (test-org-element/interpret-data):
Do not assume reverse order.
This is one of the undocumented conventions that is inconsistent with
other aspects of the parser.
* lisp/org-element.el (org-element-class):
(org-element-interpret-data): Explicitly consider anonymous syntax
node type.
* lisp/org-element.el (org-element-put-property):
(org-element-set-contents):
(org-element-secondary-p):
(org-element-adopt-elements):
(org-element-extract-element):
(org-element-insert-before):
(org-element-set-element):
(org-element-create):
(org-element-copy):
(org-element-lineage): Remove from org-element.el, using
org-element-ast function versions.
* lisp/org-element.el (org-element-headline-parser):
(org-element-inlinetask-parser):
(org-element-item-parser):
(org-element-citation-parser):
(org-element-citation-reference-parser): Assign :secondary property.
(org-element-parse-buffer):
(org-element-parse-secondary-string): Resolve deferred properties.
(org-element--cache-shift-positions): Use the new AST API.
(org-element--cache-for-removal): Use optional argument for
`org-element-set-element' to keep the needed property values.
(org-element--cache-gapless): Bump element cache version when persisting.
* testing/lisp/test-org-element.el (test-org-element/class): Update
test using the new :secondary property convention.
* lisp/ox-man.el (org-man-paragraph): Use the new AST API.
* lisp/org-element-ast.el (org-element-create): Add special case when
CHILDREN contains a single list of Org notes:
(let ((children (list a b c ...)))
(org-element-create 'type nil children))
This will simplify creating new elements when children are stored in a
list.
* lisp/org-element-ast.el (org-element-ast-map): New function
equivalent to `org-element-map', but more low-level. Unlike
`org-element-map', `org-element-ast' does not consider the details of
Org syntax - just syntax tree structure.
* lisp/org-element-ast.el (org-element--properties-mapc): New internal
helper.
(org-element-properties-resolve): New function used to resolve
deferred property values by side effect.
(org-element-properties-mapc):
(org-element-properties-map): New function to map over properties and
their values.
(org-element-resolve-deferred): New alias to resolve all the deferred
values in syntax nodes.
Major changes:
1. Property values can now deferred and computed next time when the
value is requested.
2. Some properties are stored in an array instead of plist. The
functions are inlined to turn the propery queries into direct
`aref' and `aset' calls to the plist, when applicable.
3. Secondary strings are now considered of `anonymous' type, in
backwards-compatible way.
4. New functions to map over and/or resolve deferred values of element
properties.
5. Docstrings and code consistently use "node" for generic syntax tree
elements, to not confuse the element/object terminology we use in
the parser.
* lisp/org-element-ast.el (org-element-deferred): New type user to
store deferred values.
(org-element--deferred-resolve-once):
(org-element--deferred-resolve):
(org-element--deferred-resolve-force):
(org-element--deferred-resolve-list): Helper functions used to resolve
the deferred values.
(org-element--standard-properties):
(org-element--standard-properties-idxs):
(org-element--property-idx):
(org-element--parray):
(org-element--plist-property): Store most commonly used properties in
vector for faster access. Implement inliner helpers to transform
property keywords into array indices.
(org-element-property-1): New function to retrieve property without
resolving deferred value.
(org-element--put-parray):
(org-element-put-property): Refactor, using the new property vector
when applicable.
(org-element--property): New helper function.
(org-element-property): Refactor, using the new property vector and
deferred value resolution. Add new optional arguments DFLT and
FORCE-UNDEFER. Define setters.
(org-element-set-contents): Handle anonymous nodes.
(org-element-set):
(org-element-adopt):
(org-element-extract): Rename from `org-element-set-elements',
`org-element-adopt-elements', and `org-element-extract-elements' and
keep backward-compatible alias. This is to reduce the confusion about
"node" vs. "element" vs. "object".
(org-element-create): Initialize property array correctly.
(org-element-copy): Allow copying secondary strings. Add new optional
argument KEEP-CONTENTS.
(org-element-lineage): Clarify the limitation when cache is disabled.
(org-element-type): New optional argument to identify anonymous nodes
instead of returning nil for both anonymous nodes and everything not
matching other element types.
Clarify some of the undocumented concept assumptions about Org syntax
tree that are used in Org code base. In particular, clarify that
syntax nodes are not necessarily limited by the specific Org syntax -
exporters can define new "pseudo" elements and objects.
Introduce a new concept of deferred values.
Add a new convention to keep secondary property names in a special
:secondary property.
Introduce `anonymous' element type, which was previously
indistinguishable from values that are not Org syntax elements.
Ihor Radchenko