* Initial commit
Everything is draft.
* Fix test execution
* Fix and test missing target handling
* Use the ubertransclude widget for the wikitext transclusion syntax
* Changed transclude widget in binary parser to ubertransclude
* Add a test for custom action widgets
* Don't worry about ordered attributes
The changes in 0bffae2108 mean that we don't need to explicitly maintain the ordered attributes
* Remove need to explicitly clear widget mapping variable when invoking overridden widget
* Use ts- prefix for system slot names
* Add a definition for the value widget just so that it doesn't cause errors
Of course, it doesn't actually need to be a JS widget, it could be a wikitext widget...
* Add support for positional parameters
* Ubertransclusion positional parameters should be based on name, not position
* Add support for shortcut syntax for positional transclusion parameters
* Importvariables should skip parameters widgets
* Refactor transclude widget before uberfying it
* Refactor ubertransclude functionality into transclude widget
* Replace ubertransclude widget with transclude widget
* Add wikitext shortcut for new-style function definitions
* Allow brackets to be omitted for function definitions with no parameters
* Add pragma rule for parameters declarations
* Remove erroneous "tag" property
* Add support for accessing function parameters as name/value pairs
* Be as permissive as possible with parameter names
Previously restricted to upper and lower case, digits and dash and underscore
* Rewrite some tests to use the shortcut syntaxes
* Mustn't allow commas in parameter names
* Fix crash when transcluding an undefined variable
Thanks @pmario
See https://github.com/Jermolene/TiddlyWiki5/pull/6666#issuecomment-1114692359
* Unquoted parameters should not eat a succeeding comma
Fixes#6672
* Remove extraneous code
* Allow the let widget to create variables starting with $
* Fix addAttributeToParseTreeNode handling of ordered attributes
* Reuse attribute objects when executing custom widgets
* Fix importing of function definitions
* Fix parameter handling
* Introduce genesis widget for dynamically creating widgets
See the "RedefineLet" test for a contrived example of usage
* Change tiddler separator used in wikitext tests
Underscore looked ambiguous; I kept typing dashes by accident
* Cache parse trees when transcluding variables
* Fix bug with empty strings ignored in $tw.utils.stringifyList/parseStringArray
I will pull this out into a separate PR. Fixing it doesn't cause problems for the core but I imagine it might cause issues for 3rd party code.
* Fixes to enable the transclude widget itself to be overridden
There are two big changes here:
Replace the previous "ts-wrapper" mechanism, which we had been using to redefine custom widgets inside their definitions to prevent recursive calls. Now we've got the genesis widget we can instead control recursion through a new "$remappable" attribute that allows the custom widget mechanism to be skipped.
We also extend the slot widget to allow a depth to be specified; it then reaches up by the indicated number of transclusion widgets to find the one from which it should retrieve the slot value.
* Fix genesis widget example
* Use enlist:raw to preserve duplicates
* Don't create variables with value undefined for missing parameters
* Fix variable retrieval bug with test harness
* Improve recursion detection
While retaining backwards compatibility
* Experimental support for custom filter operators
Just as we can define custom widgets we can also define custom parameterised filter operators
* Add visible transclusions component and demo
Very useful to see transclusions explicitly
Makes a good demo of a super-complicated widget override.
* Genesis widget should pass raw attributes onto child widget...
...so that it can more efficiently handle refreshing itself.
* Use consistent parse tree node property for params
* Extend transclude widget to work with old-style macros and use it for the macrocall shortcut syntax
* Clarify that the recent changes allow functions to be invoked with the double bracket syntax
In other words, the transclude widget distinguishes between functions and macros and handles the parameters appropriately
* Make the macrocall widget delegate to the transclude widget
* Switch to using \procedure to define new-style macros, and \function for custom filter operator functions
I now need to update the OP!
* Fix visible transclusion example
* Remove obsolete code
Left over after refactoring
* Better backwards compatibility for legacy recursion marker
Fixes the problem with tag dropdowns @btheado
* Fix stringifying/parsing string arrays containing newlines
A very old bug.
Fixes the ActionListOpsWidget problem @btheado
* Transclude: replace paramNames/paramValues with more robust JSON payload
More details at https://github.com/Jermolene/TiddlyWiki5/pull/6666#issuecomment-1123719153
* Rename internal "unknown" filter operator so that users cannot invoke it
* Detect recursion by tracking widget tree depth
The old recursion marker approach was very slow, and didn't catch test cases like editions/test/tiddlers/tests/data/transclude/Recursion.tid
* Use \widget for custom widget definitions, and remove need for angle brackets
Need to do some refactoring of all those isFunctionDefinition/isProcedureDefinition/isWidgetDefinition flags into a single property
* Rename <$value> widget to <$fill>
* Require $$ for custom widgets, and that overridden JS widgets must exist
See https://github.com/Jermolene/TiddlyWiki5/pull/6666#issuecomment-1133637763
* Fix invocation of JS macros
* Experimental update of the parse-tree preview visualisation
An experiment to try out using the new JSON operators for rendering the JSON parse tree that we get back from the wikify widget.
As usual with these experiments, this one is going to require quite a lot more work to finish up:
* The formatting is via direct styles rather than classes
* The formatting for attributes and properties is not yet completed
* The same thing needs to also be done to the widget tree preview
* Procedures and widgets inherit whitespace trim setting from their definition
* Missed off 22e7ec2381
* Require period prefix for custom filter operator functions
To ensure that custom filter operators cannot clash with future core operators.
* Allow custom functions to be invoked as attributes
* WIP
* Remove unneeded test tiddler
* Make is[variable] and variables[] operators resilient to fake widgets
* Fix importvariables to work with setvariables as well as set (they are aliases)
* Add support for $:/tags/Global
* Remove accidental commit 6cc99fcbe3
* Add utility function for parsing macro parameter definitions
* Introduce true global variables
The basic idea is that if we don't find a variable `foo` then we fallback to retrieving the value from the tiddler `$:/global/foo`, if it exists.
This allows us to replace the usual importvariables-based mechanism for global definitions, avoiding cluttering up the variable namespace with every macro.
In order to permit subprocedures to be overridden, we also introduce a mechanism for conditional definitions: preceding the word definition|procedure|function|widget with a + causes the definition only to occur if the specified variable doesn't already exist. In the next commit we'll apply this mechanism to the tabs macro
* Convert the tabs macro into a global
So far it appears to be totally backwards compatible... In practice, I think maybe this and the conversion of the other macros should go into a separate subsequent PR.
* Change to `?` for conditional definitions
* Fix tabs global so it doesn't crash when viewed directly
* Test showing how to un-override a core widget
* Cleaning up after f636349007
* Minor cleanups
* Clean up unknown filter
* Introduce function operator for calling functions
Can invoke any functions, not just those start with a period. And can pass zero parameters (in contrast when invoked as a custom filter operator there's no way to omit the first parameter).
* Use underscores for new system fields for global variable tiddlers
For consistency with `_canonical_uri`; unlike many system fields, the behaviour of these fields is baked into the core JS code.
* Refactor $parameters widget
The objective is to add a $depth attribute so that it is possible to reach up to retrieve the parameters of ancestor transclusions. However, doing so requires changing the encoding of parameter names so that it is not possible for a user parameter to clash with an attribute like $depth. So now we have to double up dollars on any attribute names seen by the parameters widget, just like with the transclude widget itself.
* Fix refreshing of global variables
Global variables access within attributes will automatically trigger a refresh if the attribute text changes, but that wasn't happening for transclusions.
* Remove support for $:/tags/Global
It is not needed now that we have true global variables
* Typo from f513b403fe
* Make slot fill data available to transclusions
Allows transcluded content to dynamically process <$fill> widgets within the calling transclusion
* Mark docs as v5.3.0
* Simplify metaparameters implementation
* Fix typo
* Adjust naming of transclusion metaparameter
* Fix up handling of slot/fill for custom widgets
Previously we were wrapping the body in an implicit `<$fill $name="ts-body">` widget
* Add format:json operator
I've been finding this useful for debugging, and it kind of goes with the JSON operators
* Docs: JSON operators and tweaks to genesis widget
* Docs: format:json
Also tweak to the behaviour of format:json if the input string is not valid JSON
* Fix#6721
* Revert "Fix #6721"
This reverts commit b216579255 which was committed to the wrong branch
* Fix new selection tracker to return relative coordinates
* Make use of type attribute consistent
* Docs: Transclude widget
* Simplify the fill widget
We can rely on the default processing in the base class
* Slot widget: be more defensive about negative depth values
* Parameters widget: Be defensive about negative depths
* Protect against excessively recursive functions
* FIx transcluding of functions
This first implementation concatenates the results of the filter (with no separator) and then wikifies the result.
The test in this commit is quite interesting...
* Tweak semantics of JSON operators to match #6932
This allows us to later bring in the optimisations without breaking backwards compatibility.
* Revert obsolete changes to boot.js
* Fix inadvertent whitespace change
* Remove tests related to obsolete changes to boot.js
Should have been part of 2f494ba152
* Revert changes to parse tree preview
This implementation requires #6666
* Add test to show that global widgets need not use the _parameters field
* Disable overriding core widgets in safe mode
* Coding style tweak
* More comments
* Fix caching of parse variables/macros/procedures
* Transcluded functions should operate on the entire store
* Refactor filter recursion detection to avoid an unneeded wrapper function
* Fix error in 25312b3e32
* WIP
* Revert "WIP"
This reverts commit 8654dfc679.
* When transcluding functions, pass an empty item list to the filter, and just return the first item
* Rejig genesis widget to be easier to use
* Parameters widget: protect against negative $depth
* Docs updates
* Docs updates
* Tweak comments
* Add custom view template body for globals, and a new sidebar tab under "more"
And also a custom view template title that greys out the $:/global/ part of the title
* Update function operator to return the input list if the function is missing
* Remove negation from function operator
This implementation was not useful.
* Tests and docs for function operator
* Docs tweaks
* Improve indentation
See https://github.com/Jermolene/TiddlyWiki5/pull/6666#discussion_r967655251
* Missing tests for parameters widget
* Fix visible transclude
* Docs update
* Docs typo
* Huge Documentation Update
Not quite finished, but definitely on the home stretch
* Slight optimisation to user defined widgets
* Remove implementation of $:/globals/
Performance with this implementation is inherently poor because of the need to perform a wiki lookup for each child widget created.
* Docs clarification
* Docs update
* Some widget.js cleanups
* Remove support for conditional definitions
It was introduced for use cases associated with the global mechanism that was dropped in e3d13696c8
* Docs updates
* Revert change to setwidget docs
* Docs update
* Docs updates
* Clarify/simplify some tests
* More docs updates
* Fix doc file locations
* Docs updates
* Revert modified date of docs that have only had minor tweaks
* Docs typo
https://github.com/Jermolene/TiddlyWiki5/pull/6666#discussion_r990811220
Thanks @btheado
* Transcluding functions: fix missing parameters passed as undefined
Thanks @btheado – see https://github.com/Jermolene/TiddlyWiki5/pull/6666#issuecomment-1276187372
* Parameter parenthesis should be mandatory in function/procedure/widget definitions
See https://github.com/Jermolene/TiddlyWiki5/pull/6666#issuecomment-1280404387
* Attempt to build this branch with CI
* Add release note etc for 5.3.0
* Temporary new release banner for v5.3.0
* New New Release Banner
* New test for undefined parameters
* Adjust modified times of docs tiddlers to make them easier to find
* Update release note
* Add parenthesis to the visible transclusion definition
Parenthesis were made mandatory in 5194b24108Fixes#6998
* Fix macrocall refresh issue
It turns out that this.transcludeTitle is always truthy, even if we are transcluding a variable
Fixes#7001
* Filter run prefixes should use widget.makeFakeWidgetWithVariables
* Docs typo
Thanks @twMat
* Docs: clarify function operator invocation
See discussion at https://github.com/Jermolene/TiddlyWiki5/issues/6991#issuecomment-1301703599
* Docs: Update \define pragma to cover named ends
* Docs: move tiddlers to correct directory
* Add support for named end markers for procedures, functions and widgets
* Docs note about nested macro definitions
* Rename test
* Fix detection of empty transclusions
See https://talk.tiddlywiki.org/t/exploring-default-tiddler-links-hackability-in-v5-3-0/5745/25?u=jeremyruston
* New test missed off a45349cc99
* Refactor wikified function tests
* Refactor function invocation
* Introduce new widget helper function to evaluate variables.Functions are evaluated as parameterised filter strings, macros as text with textual substitution of parameters and variables, and procedures and widgets as plain text
* Refactor the function operator and unknown operator to use the new helper
* Use the new helper to evaluate variables within filter strings, thus fixing a bug whereby functions called in such a way were being returned as plain text instead of being evaluated
* Refactor the transclude widget to use the new helper
* Update tests
* Fix positional parameters in widget.evaluateVariable()
This should clear up the remaining anomalies in #7009, let me know how you get on @btheado
* Remove 5.2.8 release note
* Fix nonstandard initialisation code for fill/parameter/slot widgets
* Update modification times of doc tiddlers
So that they are at the top of the recent tab
* Update 5.3.0 release note
* Remove custom CI step for this branch
* Restore standard sitetitle
We can now pass arrays of rule classes to the parser constructor, overriding the rules that would normally be used by the parser.
This allows us to create custom variants of the wikitext parser with their own content type.
It could also provide a basis for a new Markdown parser based on our existing wikitext parser but with new rules.
* Ignore parser rule configuration in safe mode
* Made text translatable
* Added new setting for camelcase links
* Added warning on parser rule configuration tab
See #1875
I switched this optimisation off back in
ed35d91be6, in October 2013, as part of a
big refactoring of the parsing and widget mechanism. I’ve been meaning
to switch it back on for some time.
My rough measurements suggest that this optimisation can reduce
rendering time by 5-10%.
There are still some warnings about making functions in a loop, but
I’ll fix those as a separate pull request because the fixes are more
than typographic errors.
This is a half-way through a big refactoring of the parsing and
rendering infrastructure. The main change is to separate the parse and
render trees, which makes the code a lot cleaner. The new parser isn't
yet functional enough to replace the existing parser so for the moment
you have to manually invoke it with `$tw.testNewParser()` in your
browser console. I really ought to use branches for this kind of
thing...
