Breaking release 0.25 - #5484
Open
ffreyer wants to merge 105 commits into
Open
Conversation
Collaborator
Benchmark ResultsSHA: 2a46252d5298493f1a16f247333bf0a3171ee404 Warning These results are subject to substantial noise because GitHub's CI runs on shared machines that are not ideally suited for benchmarking.
|
* move plots from elements to entries * minor cleanup * fix test failures * refactor plot extraction to be less breaking * tweak test to check `plots => Element` syntax * fix datashader again * update docs * update changelog * swap out scatter for stem to test generated legend entry * fix format * note change in legend element types --------- Co-authored-by: Simon <sdanisch@protonmail.com>
* use compute graph * get infinite bars working * allow PolarAxis to work * try to generate acceptable limits * remove unused function * fix missing else branch * allow fillto and barwidth as vectors * fix direction * fix fillto * overwrite fxaa for labels * handle log transforms explicitly * format * restore arg type check * don't need Makie. in Makie * update changelog * revert passthrough to text * add some sanity checks * formatting
* pull tweaks from #4689 * pull context validation from #4689 * handle texture atlas explicitly + fix errors * explicitly clean up shaders & programs * handle GLBuffer, GLVertexArray, Postprocessors * handle Texture explicitly * fix SSAO error * test cleanup * add some sanity checks * fix tests * cleanup finalizers, free(), unsafe_free() * skip GLFW initialized assert for CI * update changelog * fix freeing of unused uniforms & cleanup unused pattern_sections * count and check failed OpenGL cleanup * try to handle GLFW CI failure * try fix CI * require context for Texture and GLBuffer in most cases * fix incorrect uniform cleanup * CI please * fix typo * try handling GLFW init errors differently * maybe just destroy the window if the context dies? * fix typos * maybe also cleanup dead screens when trying to reopen? * tweak require_context to maybe catch error earlier * switch before requiring * avoid requiring context for get_texture and thus robj cleanup * treat scene finalizer * Treat scene finalizer in texture atlas too * add missing passthrough * move object deletion test to end * warn on dead context * avoid the last few current_context() calls * fix scatter indices * start reworking render pipeline * move gl part of Framebuffer to GLAbstraction * add show() * fix incorrect rename * change copy-to-screen to a Blit operation * fix tests * add Render step * fix SSAO * add SortPlots task and cleanup render_frame * use dedicated framebuffers * fix tests * fix depthbuffer() * simplify glDrawBuffer(s) * add prototype for abstracted RenderPipeline * switch to Vector of connections * split up multi-step stages * rename Format to avoid name collision * collect framebuffer buffers in array, prepare some utilities * integrate GLMakie with new render pipeline (prototype) * fix SSAO * fix FXAA * fix picking, use config.ssao * try recreating Framebuffer instead of removing attachments * some cleanup & performance tweaks - reuse buffers when regenerating gl render pipeline - switch Connection from Dict to Vector - inline is_compatible and promote_type - cache pipeline * try to get some debug info out of CI * fix FXAA in CI? * add changelog entry * cleanup GLAbstraction * cleanup, add extras to format, usability improvements * add stage attributes, format extras * add pipeline show and fix OIT * add GLRenderPipeline * add construct and reconstruct, unify input buffer names * use reconstruct to reuse postprocessor render objects * reuse GLRenderPipeline & be a bit more careful with caching * temp fix SSAO clearing bug * add more destroy!() to maybe fix CI * is it buffer reuse? * is it reconstruct? * Revert to green CI * Revert revert + undo debugging * add some more context switches, idk * add more context switches doesn't fix the issue but maybe helps? * cleanup enum -> name * fix plots not displaying with ssao = true in non SSAO pipeline * add test to verify all render stage parameters render * fix tick test * fix test * switch config to include (Makie) render pipeline & cover all settings * fix docs * test a few other pipelines * free more things * some performance tweaks * some more safety tweaks * add tests for Makie render Pipeline * revert fix attempts & assert correct context instead * use enum for type handling to avoid runtime dispatch * remove Connection type for better performance * cleanup rebase * minor cleanup * put finalizers behind debug constant * don't skip observable cleanup in free * cleanup require_context * fix typo * fix test errors * fix typo * rename OIT buffers * fix & test connecting two already connected nodes * add experimental GUI * remove pipeline caching * turn on ci against rebase branch * fix freeing of reused framebuffer attachments, fix tests * git pls * fix tests * improve default placement of stages in GUI * fix SSAO * add test for custom render pass * get renderpipeline to persist from activate!() * fix missing to_value * fix another incorrect free * bit of convenience * clean up * fix errors * fix freeing of GLBuffer Observables * rename Pipeline -> RenderPipeline * document rendering in Cairo, GL and WGLMakie * cleanup unsafe_free * cleanup called_from_finalizer & add comments to implied functions * some cleanup * fix more tests * merge fixes * revert ci changes * try fix docs * inline texture parameters in BufferFormat * maybe fix docs? * update tests * maybe with imports? * maybe no f,a,p? * rework output buffer handling * fix errors * cleanup some comments * move render pipeline docs to dev docs * fix docs? * does this work? * try to fix docs * update comments * remove framebuffer attachment reuse logic * merge fixes * fix changelog entry * run formatter * remove testing/experimental pipelines * update/clarify some comments * add some helper functions, fix show * test a bunch of cases for generate_buffers() * some more cleanup * rename FramebufferFactory * document some more internals * formatting * prototype depth and stencil buffers * split up the RenderPipeline file (only adds comments) * add intermediate type * prepare for MSAA * run formatter * allow unset inputs * depth without stencil works now, so use it * update tests * fix broken cleanup * format * update docs * swap order of `idx => name` in stage inputs/outputs * fix test * fix docs? * add nameless depth buffer lookup, fix test * fix docs? * fix typo * does this fix docs? * fix typo * fix show error of empty framebuffer * add flag to stop buffer reuse * add docstrings to default stage constructors * add minimal render pipeline for reference and testing purposes * show stage indices too * add mid-render colorbuffer function for debugging * cleanup scene setup a bit * don't resize buffers centrally to avoid hard coding sizes * update docs, refimg test, formatting * allow iterating GLRenderPipeline * fix test errors (missing resize) and make resizing an interface function * rename Stage -> RenderStage for consistency * rename AbstractRenderStep -> GLRenderStage for consistency * fix test * fix test * replace step with stage * Allow GLMakie render objects to render in different ways (#5436) * prototype renderobject variants just a few lines of code, nothing big * some cleanup * transition to using variants for render stages * remove compilation-only attributes from uniforms * fix name * cleanup debug prints * clean up vertex array constructor * fix a bunch of test failures * fix marker updates (broken on master btw) * fix color handling * fix empty indices in postprocessors * fix updates * restore observable tracking for render stages * fix shading compile flags * fix volume backlight * fix transparency * fix render pipeline test * break up render() function for easier reuse * simplify renderobject creation for post processors * fix undefined Screen * implement regeneration of render instructions, add test * default indices to vertex buffer length dynamically * some cleanup * avoid redundant update after init * run CI * improve hash performance of shader caching * improve mustache2replacement performance * improve type stability in shader compilation * rework pre/postrender init to maybe improve type inference, directly create strictly typed view dict * make render_in_stage type stable * cache GLSL version number * avoid unnecessary specialization * cache Symbol(string) * avoid merging uniform and buffer dicts * fix tests * formatting * partially fix type stability of default_shader * cache version string too * cache buffer glsl type names * move default prerender setup to render stages * Improve error * fix undefined normals in mesh * fix refimg test * fix screen config compat? * don't ignore directly set render_pipeline * shouldn't need closeall() for recompilation anymore --------- Co-authored-by: SimonDanisch <sdanisch@protonmail.com>
* prototype infrastructure for direction, orientation and repeated dims * simplify if block a bit * match argument_dims default to old behavior * fix typo * add infrastructure for dim_converts in axis labels * control units in tick labels from Axis * fix some errors with new Axis dim_converts * add argument_dim_kwargs * run add_attributes first so that dim_converts can include them * fix/update hvlines * get heatmap, surface, contour, contourf working * update hist * handle image * remove dynamic unit prefixes * update band, errorbars, rangebars, hspan, vspan * update remaining basic_recipes plots that should work * update stats recipes * diy missing dim_convert discard * cleanup print * avoid errors for empty rich text * fix label typing * fix voronoi 3 args, update timeseries * add tests * fix boxplot, 3D contour * fix 3D surface convert, test remaining primitives * fix log units * remove unused code * fix triplot/poly conversion * revert deleted convert_arguments but remove recursion * fix empty rich text properly this was easier than I though * fix qqplot/qqnorm/early convert kwargs application * fix "]" from formatter ending up in superscript after a "^" * fix load order * rename attribute + docstrings * update Axis3 * add long units * fix Int error in Date resampling * fix/update tests * revert to late attribute adding and only add required attributes early instead * fix tests missing parent_is_scene * revert unnecessary change to add_convert_kwargs!() * fix PlotSpec * remove unused `try_dim_convert` and `should_dim_convert` (keep name for compat) * drop unnecessary method * update docs * run formatter * minor cleanup * get point-like data working * remove prints * explicitly exclude Axis3D plots * allow ranges in argument_dims() * update more point like arguments * fix getindex(nothing, idx) * fix failing dim converts tests * fix empty array application * fix convert_arguments tests * fix tooltip * fix annotations * fix tooltip * make `force_dimconverts` settable to fix Axis3 * fix mesh * fix tests * fix docs * avoid converting non data space * improve documentation of conversion logic and steps * add refimg for label attributes * avoid need for ::Automatic in show_dim_convert_in_...() * allow early conversions when dim converts are forced (+ some restructuring) * add tests for point-like data & fix conversions * run formatter * fix and test band with vectypes * fix docs? * default units to axis labels only * default multiple point-like arguments * make sure every recipe has dim convert exit paths * some cleanup + docstrings * fix dodgy docs * add section about conversion target types to recipe docs * update changelog * format * fix typos * fix multihist vector of vector args * fix tests
* chore(changelog): Update CHANGELOG.md * test: Add failing test for issue #1939 * fix!: `merge!` precedence and attribute precedence fixes - Update `merge!` function to align with the right precedence which is the standard for `Base.merge!` - Add `leftmerge!` function for updating immutable struct attribute fields, where shadowing cannot be used - Reorder the arguments where the original wrong precedence `merge!` was used to use the correct precedence `merge!` implemented in this commit without change in output - Use `merge!` instead of `merge` in `plot_attributes` function to avoid `deepcopy` and improve performance - Fix precedence in `default_attribute` function BREAKING: `Base.merge` and `Base.merge!` now has the opposite precedence for `Attributes` arguments * perf: Remove the check for the existence of the key in target The key is present in the branch because, we already checked for the opposite in a previous branch. * fix: Remove the unused `merge_attributes!` function --------- Co-authored-by: Frederic Freyer <frederic481994@hotmail.de>
* Blocks as complex recipes * fix incomplete init due to notify * support externally defined Block * support indexing and on-demand layout generation * support auto axis via `plot(block[i, j], ...)` * add lazy `block.blocks` * add plots tracking * add some documentation * ignore undefined block.layout for now * fix block deletion errors * avoid .val perfectly safe, surely * remove/fix more notify's * fix dim_converts * fix some more issues * fix Dates * fix block cleanup * fix textbox size * fix tick updates for unit changes * fix unit tick updates with dispatch instead * fix infinite loop in menu * fix IntervalSlider size * fix Axis3 limits * fix textbox realigned on delete & try to clarify code * fix html widgets ? * remove unused plot type from _register_input_arguments!() * implement argument conversions for blocks * make notify(computed) actually do something * auto-export blocks * fix code order * export `@Block` * auto figure for blocks * bring back attribute types inits * format * add changelog entry * fix error * fix tests * fix docs error? * split argument handling to make it easier to reuse * add some convenience for map! and register_computation! * update docs * fix Axis in complex recipe block not initializing correctly * fix compute errors again * don't type restrict cycled recipe attributes * fix poly strokecolor type changes * fix box strokecolor type changes * add `update!(block, ...)` * allow (xs, function) conversions in blocks * clean up error * add test for complex recipe + block updates * ok maybe not repr() * format * fix WGLMakie * fix type change error in CairoMakie heatmap (Range <-> Vector) * fix error * simplify code * remove block.plots * don't allow layout to be added after init and improve show() * reorganize code, fix attribute autocomplete, document attribute converts * fix docs * set up SpecApi in Blocks * add getindex and setindex for block arguments * allow switching between BlockSpec and GridLayoutSpec * add test for SpecApi in blocks * format * allow empty/no attributes and no begin ... end block * allow convert_arguments to not return a spec *tuple* * fix nested GridLayout deletion in SpecApi (+ nicer show methods) * add `Container` block as an implementation for `Block(spec)` and GridLayout wrapper * fix autolimits (same value update does not trigger limit update) * update SpecApi docs * format, fix docs? * add a couple tests * avoid double update * add docs example as a refimg test * fix update before display * rename FigureAxis to FigureBlock (with fallback) * fix getindex with side * fix format * add first/lastindex * fix merge * more thorough Observable compat to fix Axis3 updating & other potential issues * format * fix Union conversions * fix ticklabelspace update * fix legend layout update being one step behind * fix incomplete layout of Axis * fix Textbox not reacting to placeholder * fix Menu hover/inactive color not triggering updates * rework test to include updateless reference * update changelog * format * try fix test inconsistencies * try sleep * add tick events to CairoMakie so Toggle animation can finish * fix typo, reduce sleep time
* experiment with simulated nesting * don't reuse the key array so graph views can be accessed multiple times * adjust function chain so values can cause nesting * convert Attributes() in inputs to nesting * update old axis3d! * remove print * fix per-segment linewidths not working * fix scaling * fix static value access in axis3d how did this even work? * fix nested setindex/setproperty * fix tests * add compat for `plot.attributes[].entry` * add some tests * improve some errors, printing * improve interfacing with nested nodes + tests + docs * fix format * add changelog entry * add iterate for ComputeGraphView * format * allow passthrough of nested nodes and ComputeGraphViews + test * format * restore model, transform_func filtering * fix variable name * get nested merging to work * did I mention Observables make things more difficult? * format * add more careful indexing for old axis * add sizes to size missmatch warning in refimages * clean up merge_without_obs * add a way to create empty nesting layers * fix and test add_input!() with callback * fix nested update!(), test updating * fix toggle anim in test * rename to_graph -> root * use local name for callbacks instead of the full name * Revert "use local name for callbacks instead of the full name" This reverts commit 538de00. * fix and test plot attributes from theme * format
* refactor LineAxis, Axis to ComputeGraph * don't run initialize_block! twice in precompile * deal with nothing * fix some stuff * run benchmark too * fix recursive Observable updates discarding notifies * more locks for safety * fix GLMakie not updating on just plot insertion * Add set_type!() * add flag to force updates to propagate in Inputs * add ExplicitUpdate wrapper to overwrite update propagation behavior * improve some errors * allow Block attribute init to be extended * split up scenearea node so it's usable with computations * move limits to compute graph * minor test fixes * guard against infinite recursion in axislinks * fix incorrect bbox in protrusions * fix dim converts * fix linked layout spec not initializing links * fix axis label translation * fix recursion loop in Colorbar protrusion * minor test cleanup * default linked axes in SpecApi to limit union * fix autolimits with linked axes (same value reset) * remove not yet defined type * fix type to Float64 * fix type in test * fix some things that were probably wrong * reorder code to skip an Observable * cleanup leftover Observables in LineAxis * more LineAxis cleanup, update tight_ticklabel_spacing!() * avoid checking boundingboxes for each dimension * fix 0 width limits limits * allow mark_dirty!() to exit early by having mark_resolved!() resolve parents * avoid recursive lock, use `@lock` * move old reset_limits to Axis3 * fix autolimits, same value reset in Axis3 * fix `set_type!()` for inputs * fix PolarAxis autolimits/same value limit resets * minor cleanup * benchmark against breaking * convenience + docstrings for forced_update and ExplicitUpdate * test forced_update, ExplicitUpdate, set_type!() * improve and test recursively connected Observable handling * format * restore xlims!()/ylims!()/limits!() setting linked axis limits * don't trigger two updates when calling limits!() * fix limit errors * more compute docs * update changelog * fix categorical update tests? * more sleep for WGLMakie * turn mark_dirty skips back on * try to optimize only the init -> set_type change for attributes seems to be relevant... * fix state corruption if mark_dirty!() runs during multi-graph resolve adjusted from DataInspector pr * make directly calling mark_dirty!() safer * graphs can be connected in a loop * move some of the nesting code * make mark_dirty!() irrelevant to resolve and refactor input passthrough to go through a single computation (1) should fix multi-graph state desync (2) should fix multi-graph value desync maybe without destroying WGLMakie this time? * WGLMakie bandaid? * give up and just lock globally * fix error printing * remove input bus * format * fix sanity checks * revert benchmark base
* try documenting PointBased arguments in a reusable way
* Add utils for getting applicable methods + docs
* update basic_plots argument docs
* update some more docstrings
* update band, barplot, bracket, contourf, contour, contour3d
* update datashader, errorbar, rangebar, hvlines, hvspan
* I forgot lists needed indention
* refactor argument_docs() helper
* update pie, raincloud, scatterlines, spy, stairs
* switch to plural by default
* rest of the basic_recipes folder
* update stats plots
* why does this error?
* revert failing use of conversion_trait
* fix typo
* simplify code a bit
* recover conversion_docs function
* add some warnings
* run formatter
* fix missing `; attributes`
* fix indention
* tweak wording
* first prototype
* fixes
* prototype for generating docs completely from the repl docs
* small fixes
* restructure docs
* cleanup
* move plot and attribute examples to proper place and use normal docstring for argument docs
* fix doc build
* fix no docstring case
* fix user docs
* small tweaks
* fix argument docs and bring it into one format
* make runic happy
* fix contour examples
* tweaks and update all remaining plots to new format
* fix tests and more tweaks
* I guess we dont return a figure
* bit of cleanup
* simplify `help()` function
* refer to help() or `?Plot.attribute`
* add attribute grouping
* why no docs? plots bad?
* fix errors
* fix missing refs?
* re-add overview.md, enable all plots except timeseries
* add grouping for online docs
* move generic arrows docs into 2d and 3d variants
* add autocomplete for `?Plot.attribute`
* minor cleanup for examples
* rename examples header for plot docs
* add output variable for SpecAPi call signature
* add linebreaks to keep argument docs in the observable universe
* clean up some docstrings (description, arguments, examples, attribute groups)
* clean up some more
* cleanup stats plots
* cleanup basic plots/primitives
* restore scatter online docs
* make docstrings more adjustable, add documentation
* update changelog
* formatting
* allow general note in last item
* fix arg docs for `Tuple{...} where ...` types
* add docstring for types_for_plot_arguments
* tighten VecTypes arguments to Points to reduce clutter in docs
* replace var"#..." with their types in type docs
* fix textlabel conversions
* avoid another VecTypesVector target
---------
Co-authored-by: ffreyer <frederic481994@hotmail.de>
5 tasks
Contributor
|
I can tell you, that it is breaking for us in DimensionalData, but it is on us, because we depend on Makie.attribute_names. I came upon this PR while trying to understand what changed and how I might be able to fix it. |
felixcremer
reviewed
Jul 16, 2026
5 tasks
Contributor
|
Howdy, I forgot to qualify these in the DQ extension: Makie.jl/Makie/ext/MakieDynamicQuantitiesExt.jl Lines 25 to 27 in cbb3c6c - needs_tick_update_observable(conversion::M.DQConversion) = conversion.quantity
- show_dim_convert_in_ticklabel(::M.DQConversion) = false
- show_dim_convert_in_axis_label(::M.DQConversion) = true
+ M.needs_tick_update_observable(conversion::M.DQConversion) = conversion.quantity
+ M.show_dim_convert_in_ticklabel(::M.DQConversion) = false
+ M.show_dim_convert_in_axis_label(::M.DQConversion) = trueIs there any chance that this could make it into this PR? I didn't notice the issue until I saw that things like this weren't working as expected: fig = Figure()
ax = Axis(fig[1, 1])
lines!(ax, (1:10)u"m") # Unit suffix is empty
fig |
This was referenced Jul 30, 2026
…#5726) Post reference image status via workflow_run The review-page job cannot post commit statuses for fork PRs because their token is read-only. Upload the status data as an artifact instead and post the status from a workflow_run-chained workflow that runs in the base repository's context once CI completes. The review page's artifact id is passed through the status data because upload-artifact with archive: false registers the artifact under the file name (index.html), so a name lookup would not find it. The chained workflow validates all fork-controlled fields before use and never checks out PR code.
Collaborator
Author
Yup, done |
Collaborator
Author
|
/benchmark 20 |
Collaborator
Benchmark ResultsSHA: 26891cea60dab5dd7ea5775d46db742e0d71e30d · 20 samples Warning These results are subject to substantial noise because GitHub's CI runs on shared machines that are not ideally suited for benchmarking.
|
Collaborator
Author
|
/benchmark 20 |
Collaborator
Benchmark ResultsSHA: 229530bc6559bd2f0bddc222fb9e550f5515de5c · 20 samples Warning These results are subject to substantial noise because GitHub's CI runs on shared machines that are not ideally suited for benchmarking.
|
* simplify register_computation! to map! * add `@nospecialize` to register_computation! callbacks * fix error deprecation in tests * other cleanup * fix benchmark target not defaulting to pr base branch * fix FieldError
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
8 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.
Description
Breaking PRs to merge: (Some of these may not need to be part of a breaking release. I'm rather quick to call something breaking)
block.attr[] = block.attr[]andsetfield!(block.attr = ...). May contain update issues due to equal-value discards)units_in_labelto axes, removes autoscaling of units, changes when dim converts apply)show_datainterface, deprecatesinspector_hoverandinspector_clearattributes (TODO?))legendelements(::Plot, legend)no longer sets plots as they were moved out of LegendElement structs)shared_attributes, updating, infrastructure) #5620 (rework attributes infrastructure to handle nested attributes (plot and block recipes, initialization, docs, updates))mergeprecedence fix forAttributes#5332 (if we want to fix it we should probably do it in a breaking release?)skip_updateas a replacement for returningnothingin ComputePipeline callbacks #5754Consider:
ifelsebranches in ComputeGraph edges #5749 (non-breaking)PRs to merge outside of this:
add interactive zoom inset #5456Fixes:
barplotcontains gap when usingyscale=log#4549hist(dates)MethodError #3889, fixeshistshould work on categorical inputs #4409meshscatterlinesdoes not work with Point2 with DateTime #4774, fixes Sometimes units are not handled #3945base_unit(::Float64)#4085, fixes plot() returns error when called on data of Vector{Quantity{Int}} due to arguments passed to Unitful.Quantity() #5231. Well sort of, the input data is aVector{Quantity{Float64}}which now successfully gets converted as a unitless unit. Generates and empty label suffix though[]scatter!fails forVector{Union{missing, Float64}}#5089, maybe because earlyconvert_arguments()run more consistently?scatterlines.inspector_labelis broken #4713access to undefined referencewith child plots #5383 by avoiding the need to setinspector_labelfor child plotsvlines#5520mergeprecedence fix forAttributes#5332mergeonAttributesdoes not work in the same way as the one in the standard library #1939Legend(..., ax)to update with plots created after Legend #5672axislegendcould update when new plots are added to the axis #3207TODO:
x/y/zlims!()still work the same in all axis types, specifically after interactively changing themconsider making export skipping in recipes user facing, i.e. add documentation and improve syntax (e.g. kwarg-likeadded to master by Juliusexport = true/false)@forwarded_layoutfix orremoveshort_unit, rework suffix to work with different label string typesChecklist