Conversation
Update the version to 2.21 and the copyright years to include 2026.
…ection-plus Harden system calls plus.
This is a result of a workaround for a bug in the JSXGraph library and failure to properly convert all of the code in the `keyDownListener` function to work outside of the library. The `Const` object is not defined. To access those constants instead the `JXG` global object must be used.
Fix a minor bug in the GraphTool.
When collecting a matrix array answer use `student_array` instead of `student_formula` as `student_formula` should always be a formula, and this could fail to convert to a formula if an answer blank is left empty, causing a future `typeMatch` error. Fixes #1406.
When checking if `$other` is an object, check that it is actually a `Value` object instead just that it is a reference in the various `typeMatch` methods. This is done by using `Value::isValue($other)` instead of `ref($other)`. In addition add a test for `Value::isValue($other)` in the `Value::Formula::typeMatch` method to be consistent with the other methods. This was suggested by @dpvc.
Don't use student_formula to collect matrix array and update typeMatch methods.
This fixes issue #1349 and is the result of the discussion in that issue. See the issue for details and example problems.
This color has a contrast ratio of 6.68 against the white background. The red color has a contrast ration of 4.00 which is not sufficient for accessibility purposes.
The method returns an array containing references to arrays that form a
partition the vertex indices into the connected components of the graph.
For example, for the graph with vertices E, F, G, H, I, J, K, and L, and
edge set {{{E, L}, {F, G}, {F, L}, {G, J}, {H, I}, {J, K}, {J, L}}}, the
method will return ([E, F, G, J, K, L], [H, I]).
Instead of trying to place the labels using the coordinates, use the `anchor` and padding options to get better positioning. The primary advantage is now the labels don't float away when the image is enlarged. This was not done originally because at the time this macro was implemented the `anchor` and `padding` options didn't exist. The weights for the default layout and the wheel layout are still positioned along the perpendicular vector for now. The problem is that those labels are rotated, and that does not work well with the anchor. This is a TikZ issue (I implemented the anchor for JSXGraph to work the same as the TikZ anchor option). The problem is that the rotation is around the position of the anchor, and not around the center of the text. The usual solution for this in TikZ is to use a `\rotatebox` on the node contents. Perhaps another rotation option could be addded to the plots macro that would rotate the text instead of rotating around the anchor position.
an invalid edge set for a graph with a single edge.
First, the method is now quite a bit more efficient. Rather than finding and sorting only the weights of the edges in the graph, and then searching through the graph to find those edges at each step in the algorithm, the edges with the weights are all listed and sorted by weight (more like the actual sorted edges algorithm works). So there is no need to find the edge later, you just follow the algorithm and process the edges in order. Second, the return value of the method is reordered and more data returned. See the updated POD for good documentation on what is returned. This makes the method return a lot more useful information that can be used for constructing a solution to problems using the sorted edges algorithm.
First, the method is now quite a bit more efficient. Rather than finding and sorting only the weights of the edges in the graph, and then searching through the graph to find those edges at each step in the algorithm, the edges with the weights are all listed and sorted by weight (more like the actual sorted edges algorithm works). So there is no need to find the edge later, you just follow the algorithm and process the edges in order. Also make the algorithm terminate once the minimal spanning tree is complete as it should. Second, the return value of the method returns more data. See the updated POD for good documentation on what is returned. This makes the method return more useful information that can be used for constructing a solution to problems using the sorted edges algorithm. These is the basically the same changes that were made for the `sortedEdgesPath` method.
This returns a translated string description of the graph. The string that is returned is suitable for use as the "alt" text for the image returned by one of the image methods.
This is giving warnings when the unit tests are run.
Make list answers work in MultiAnswer and RadioMultiAnswer part answers.
Fix a typo in the `ProhibitBeginproblem.pm` PG critic policy.
Fix an ambibous `abs` call in `AnswerChecker.pm`.
Using `//=` in signatures is a feature that is not available until version 5.28 of perl. I am not entirely certain of why the added `to_string` is needed or why it works with newer versions of perl, but it causes an issue with older perl versions as well. This fixes issue openwebwork/webwork2#2978.
This addes a comment to the POD about the `scores` array that is set internally, and that can be used in a custom checker.
Add a note in the `parserMultiAnswer.pl` macro about the `scores` array.
Pg critic fixes for older perl
Some improvements and additions to the `SimpleGraph.pl` macro.
This adds two options to plots, `texPackages` and `tikzLibraries` to Plots, that can be used to add additional packages and libraries fro generating TikZ output. This also adds "amsmath" in the list of default packages that are loaded.
This is for problmes that use the `dragndrop.js` JavaScript and the `DragNDrop.pm` module via the `draggableProof.pl` and `draggableSubsets.pl` macros. The elements in a drag and drop list can now be focused using tab and shift-tab. Once focused the arrow keys move the elements around. If an element is moved with the keyboard controls the changes are announced in a visually hidden span. Note that drag and drop actions via the mouse cursor are not aria announced for now. I am assuming that a screen reader user would not be using the mouse for drag and drop. There is now also a "Drag and Drop Help" button that is shown below the drag and drop lists. If pressed, help is shown describing the keyboard controls. This help can be customized by the problem. There are new options for the macros that allow for customizing the help and associated texts. See the updated POD in the module and macros for details. Also, I don't like that there are some options that are for the `DragNDrop.pm` package that are passed in from the macros, but are different in the macros than in the module. Basically the first letter is upper case in the macros, but lower case in the module. So the upper case first letter variants is deprecated (but will still work with a compatibility layer). I don't like that the macros use Pascal case to begin with for options. Options should be camel case. Note that I also made all of the texts for the drag and drop object translatable via `maketext`. The texts can be customized per problem as mentioned above, but if only a translation is needed there is no need for such customization.
This sets the `aria` option from the JSXGraph library for all graphed objects. Thus objects are described as they are graphed. I am sure this will still need some work, but this makes the graphtool at least basically accessible. There is also a need to specifically set the tabindex for objects (usually to the empty string which is the closest thing to unsetting it that the JSXGraph library allows). Otherwise JSXGraph now sets that to -1 for everything and tries to focus those things with javascript and that is interfering with the graphtools own focus behavior. This is more of what was done in #1357. Note that the aria labels are not translated. Although nothing for the graphtool is. The usual data attribute approach really would be a mess for this. There are a lot of strings. Translating the graphtool would take a lot of effort with the current methods for javascript translation.
Add better documentation of the convert-to-pgml script. Separate the loadMacros to a separate subroutine and perform error handling in the subroutine to better capture possible syntax errors. Also, add a better way to parse the answer blanks in PGML form.
And rewrite the list of macros to be removed from the list.
The loadMacros parsing is now much more versatile. In addition to basic single and double quoting, and `qw` quotes, it can handle any quotelike quotes. There is a minimal attempt to keep comments inside of a `loadMacros` call. Basically, if a macro is removed from the list, then a comment after it is also removed. If a macro is kept, then so is the comment that was after it. If there is a line that only has a comment on it, then if the preceding line had a comment it is assumed that the comment continues that one, and if that is kept so is the next line. Otherwise the comment line is dropped. Previously the `MathObjects.pl` macro was removed if loaded, since it is loaded by `PGML.pl`. Now `niceTables.pl` is also removed, since `PGML.pl` also loads that. The naive approach of replacing `ans_rule` calls with PGML has been removed. Instead any `ans_rule` call is just wrapped in `[@ ... @]*`, and the corresponding `ANS` calls are left alone. One advantage of this approach is that after the conversion is completed, the problem actually works. In addition, the previous approach was clobbering the `ans_rule` calls in many cases that give information about what needs to be done to properly convert the answers into the PGML answer syntax. The con is that the conversion into PGML syntax must be done manually. However, since that really was the case before other than basically just putting a non-functional PGML answer template into place, that is not really a tradeoff. Instead of handling only two very specific forms of heredoc/nowdoc syntax `SOLUTION(EV3(<<'END_SOLUTION'));` and `TEXT(EV2(<<EOT))` where the heredoc/nowdoc terminator was fixed and must be exactly what is shown above, any such form is handled. The heredoc/nowdoc terminator can be anything the author specified and an unquoted, singled quoted, or double quoted terminator can be used. In addition both `EV2` and `EV3` are covered, and the `HINT` form is also handled. In addition spaces can be in those constructs as allowed by Perl. Note that there are OPL problems that do all of those things, so it was really quite naive to only literally cover those two cases. Variable interpolation now actually works much better, including for hash and array access, and hash reference and array reference access. Note that a file that already uses PGML and does not have any of the old style PG constructs is actually completely left alone. @pstaapb's #1425 still modifies those files in some cases. The `bin/convert-to-pgml.pl` script no longer accepts `-s|--suffix` argument. Instead it accepts a `-e|--extension-prefix` argument. That is much the same except that its value is prepended to `.pg` instead of used as the extension itself. It defaults to `pgml` as the `suffix` option did before. So if `problem.pg` is converted, then the result will be written to `problem.pgml.pg`. Thus the resulting file has the correct pg problem file extension. A new option for the `bin/convert-to-pgml.pl`, `-p|--output-path` was added. When given, the converted file is written into that directory with the same name as the original file. The `-b|--backup` and `-e|--extension-prefix` options are ignored, unless the output path resolves to the directory containing the original file, in which case those options will still apply. TRhere were some other minor things fixed along the way as well that I did not list here, such as an issue with the conversion of the `$HR` variable into a PGML `---` in which the `---` was added, but the `$HR` variable left and changed into `[$HR]` resulting in ``` --- [$HR] ``` in the converted file. This was done in collaboration with Claude.
This is something that has been missed with MathQuill inputs for a while. The original text input has an `aria-label`, but when it is hidden that becomes meaningless. So transfer that label to the MathQuill input. The input also needs the `role` to be set to `textbox` for this to work.
1. The DIV containing the standard answer boxes are set to use dir='ltr' so that then result/feedback icon appears to their right even when a answer box is used directly in the text in a problem which is being rendered in an RTL setting. This will match the behavior which occurs when answers are inside an LTR span in order to handle interleaved math and answers in RTL language problems. It should have no visible effect on problems in English or other LTR languages, but makes things consistent for problems in RTL languages. This change was suggested to me by @drgrice1. 2. Set the CSS for the feedback message box to have a default setting of dir='ltr'. This will be changed automatically to dir='rtl' when the course language is either set to Hebrew or Arabic (as those languages are set to use HTML with dir='rtl' set, so will use the RTL version of the CSS.) That will make feedback messages in such courses (which are expected to mostly be in the RTL language) render in the appropriate RTL manner. There should be no effect on courses in LTR languages.
…lnerabilities. Actually only `postcss` needed to be updated, but I updated all of the development dependencies to their latest versions (only `sass` and `yargs` were not at their newest versions). The development dependencies are easy to test. If the `generate-assets.js` script still works as it should then all is good. Note that you can test by executing `npm ci` since that runs the script. I left production dependencies as they are. I didn't actually even check if there are any upgrades, but those can take more time to properly test since we actually use those.
Update the npm development dependencies once again to fix security vulnerabilities.
2 small changes to improve behavior in RTL problems/courses
Transfer aria labels of text inputs to MathQuill inputs.
Move unionTables.pl to deprecated folder
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Pgml convert improvements revised
When creating a copy of a context, only keys on the new context were copied, instead all keys from the old context should be copied. This is fixed by looping over `$self` instead of `$context`. This fix was suggested by @dpvc.
tex, string, and separator options for units
Sig fig context
Fix bug in copying a context.
change Round function to use sprintf and combat machine rounding error
somiaj
approved these changes
Aug 4, 2026
Alex-Jordan
approved these changes
Aug 4, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
7 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
This is the release candidate for WeBWorK 2.21. Please re-target any pull requests that you want to get into the release for this branch.