---
tags: Animation, Animation 2025
---

# Animation Filtering Replacement

[TOC]

## Animation Data Sources

Current datasources (and a few desired ones):

* FCurves in Actions on:
* Object (regular action, but also pose data)
* [color=purple] Collection (targeting contained datablocks)
* [color=purple] Scene (targeting contained datablocks)
* Shapekey (sort of object-owned, but also heavily mesh-related data)
* Object Data (Armature, Camera, Curves, Grease Pencil, Lattice, Light, LightProbe, Mesh, MetaBall, PointCloud, Speaker, Volume)
* Other animatable IDs (CacheFile, FreestyleLineStyle, Mask, Material, MovieClip, NodeTree, ParticleSettings, Scene, Simulation, Sound, Tex, World)
* Grease Pencil frames
* Masks
* Drivers
* NLA
* Tracks
* Strips
* [color=purple] Non-numerical data, like camera switches

Purple tag: yet to be implemented.

## UI Filters / Source Selectors

We may want to split these up. Currently these are about the source of the shown data, while others are filtering that data. I’m listing them here in one place anyway, because they are currently rather entangled.

* Unfiltered (so everything in the file, except unused Actions)
* Selected object / bone
* Hidden (by the user)
* Errors (animation data for non-existent properties)
* Specific Action
* Specific Source (Shape Key, Grease Pencil, Mask, Cache File)
* [color=purple] Selected properties (and multiple, configurable pages of those)
* [color=purple] Specific property types (only transform, only custom props, etc)

Purple tag: yet to be implemented.

### Code-side Filters

These filters are currently defined in Blender’s code to achieve the above, and usually combinations are used to query for animation data.
| Filter | Expanation |
|---------------------------------|---------------------------------------------------------------------------------------------------------------------------------|
| `ANIMFILTER_DATA_VISIBLE` | Data which channel represents is fits the dope-sheet filters (i.e. scene visibility criteria) |
| `ANIMFILTER_LIST_VISIBLE` | channel is visible within the channel-list hierarchy (i.e. F-Curves within Groups in ActEdit) |
| `ANIMFILTER_CURVE_VISIBLE` | channel has specifically been tagged as visible in Graph Editor (Graph Editor Only) |
| `ANIMFILTER_LIST_CHANNELS` | include summary channels and "expanders" (for drawing/mouse-selection in channel list) |
| `ANIMFILTER_ACTIVE` | for its type, channel should be "active" one |
| `ANIMFILTER_ACTGROUPED` | channel is a child of the active group (Actions specialty) |
| `ANIMFILTER_SEL` | channel must be selected/not-selected, but both must not be set together |
| `ANIMFILTER_UNSEL` | |
| `ANIMFILTER_FOREDIT` | editability status - must be editable to be included |
| `ANIMFILTER_SELEDIT` | only selected animchannels should be considerable as editable - mainly for Graph Editor's option for keys on select curves only |
| `ANIMFILTER_ANIMDATA` | Flags used to enforce certain data types. The ones for curves and NLA tracks were redundant and have been removed for now. |
| `ANIMFILTER_NODUPLIS` | duplicate entries for animation data attached to multi-user blocks must not occur |
| `ANIMFILTER_FCURVESONLY` | avoid channel that does not have any F-curve data |
| `ANIMFILTER_TMP_PEEK` | for checking if we should keep some collapsed channel around (internal use only!) |
| `ANIMFILTER_TMP_IGNORE_ONLYSEL` | Ignore ONLYSEL flag from #bDopeSheet.filterflag (internal use only!) |

These combine the following concepts:
* Visibility as per global toggles in the user interface.
* Properties of animation channels (like FCurve FCURVE_VISIBLE, FCURVE_SELECTED, etc.)
* Relationship with other animation data
* Animation source properties (intrinsincly editable vs. non-editable)
* Properties of the animation data source (number of users)
* Animation data type (FCurve, Grease Pencil)
* “Internal use only”

### Uses of those filters

- Transform code (`transform_convert_nla.c`, `transform_convert_graph.c`, `transform_convert_action.c`)
- NLA code (`nla_buttons.c`, `nla_channels.c`, `nla_select.c`, `nla_edit.c`, `nla_draw.c`):
- which data to 'push down'
- which NLA tracks are selected & visible (in two different ways, with and without `ANIMFILTER_ANIMDATA`)?
- what are the active strip/track/animdata?
- which NLA tracks & strips are draw where on screen? Various incarnations of the same query across different files.
-
- Animation channels (`anim_channels_define.c`, `anim_channels_edit.c`)
- Mostly just 'what is visible?' and 'what is selected?', with or without `ANIMFILTER_NODUPLIS`, and depending on whether editing is necessary, `ANIMFILTER_FOREDIT`. Rather inconsistent with the other flags, though.
- Selection
- Selection / 'activeness' state synchronisation (`anim_deps.c`)
- Key selection (`graph_select.c`)
- The actual filtering code (`anim_filters.c`)
- Keyframes (`keyframes_edit.c`, `keyframes_keylist.cc`)
- Only `ANIMFILTER_DATA_VISIBLE` with the occasional `ANIMFILTER_FCURVESONLY`
- Grease Pencil (`editaction_gpencil.c`): copy-pasting of GP frames.
- Context (`screen_context.c`)
- Selecting driver channels (`screen_ops.c`)
- Animation editor drawing and interaction (`action_draw.cc`, `action_edit.cc`, `graph_draw.c`, `graph_ops.c`, `space_graph.c`)
- Animation data manipulation (`action_edit.cc`, `graph_edit.c`, `graph_slider_ops.c`)
- Baking to 'ghost curves' (`graph_view.c`)
- Animation data queries (`action_select.cc`, `graph_view.c`), both screen-space and the regular 'selected and visible' queries.
- Getting active FCurve, some poll functions (`graph_util.c`)

### Almost name collision

:::success
What's described below is handled in [#106028: Refactor: remove enum eAnimFilterFlags](https://projects.blender.org/blender/blender/pulls/106028).
:::

The above flags are defined in `ED_anim_api.h` as `enum eAnimFilter_Flags`. BUT: in `ED_keyframing.h` there is `enum eAnimFilterFlags`. Note the different names, one has `_` and the other does not. The latter has the following flags:

| Filter | Expanation | Remarks |
| ---------------------- | ---------------------------------------- | - |
| ANIMFILTER_KEYS_LOCAL | only include locally available anim data | here 'local' means 'on the datablock' and not 'its materials/shapekeys' |
| ANIMFILTER_KEYS_MUTED | include muted elements | never set, so can be removed |
| ANIMFILTER_KEYS_ACTIVE | only include active-subelements | never set, so can be removed |
| ANIMFILTER_KEYS_NOMAT | don't include material keyframes | never set, so can be removed |
| ANIMFILTER_KEYS_NOSKEY | don't include shape keys (for geometry) | never set, so can be removed |

These are used in:

- 3D Viewport drawing (`view3d_draw.cc`), to colour the name of the selected object depending on whether there is a key on the current frame or not. Always uses `ANIMFILTER_KEYS_LOCAL`.
- Auto-keying code (`keyframing.c`), to determine whether the 'only replace existing keys' mode should auto-key. Always uses `ANIMFILTER_KEYS_LOCAL`.
- `fcurve_frame_has_keyframe(..., filter)` is used in a few places, sometimes with, sometimes without filters. But the only filter it uses is `ANIMFILTER_KEYS_MUTED`, which is never set.

Conclusion: the use of these filter flags seems to have been reduced over time to only, always use `ANIMFILTER_KEYS_LOCAL`. This means that:

- the entire set of flags can be removed,
- the code that handles those filters can be removed, and
- the `filter` parameters can be removed from the functions that pass them.

## Editor-side Behaviour

Editors should be able to set up their own sources and filtering. This should allow pinning of animation data to those editors, making them independent of the selected/active datablocks in the scene.

## Transform System Integration

It should be possible to connect the filtered data with the transform system.

## Dependency Graph Integration

Animation Data Sources need to be able to set up depsgraph relations.

## Data Flow Diagram

erDiagram
    ID ||..o| ActionDataSource : "has"
    ID ||..o| NLADataSource : "has"
    OBJECT ||..o| GreasePencilDataSource : "has"

    ActionDataSource ||..o{ ChannelGroup : "produces"
    NLADataSource ||..|| NLARetimer : "has"
    NLARetimer ||..o{ ChannelGroup : "produces"
    GreasePencilDataSource ||..o{ ChannelGroup : "produces"

    ChannelGroup ||..o{ Channel : "contains"
    ChannelGroup ||..o{ ChannelGroup : "contains"

<!-- ### Example graphs for reference

Just for reference on how to draw these things.

digraph summary{
		start [label="Start with a Node"]
		next [label="Choose your shape", shape=box]
		warning [label="Don't go overboard", color=Blue, fontcolor=Red,fontsize=24,style=filled, fillcolor=green,shape=octagon]
		end [label="Draw your graph!", shape=box, style=filled, fillcolor=yellow]

		start->next
		start->warning 
		next->end [label="Getting Better...", fontcolor=darkblue]

}

digraph structs {
	node[shape=record]
	struct1 [label="<f0> left|<f1> mid\ dle|<f2> right"];
	struct2 [label="{<f0> one|<f1> two\n\n\n}" shape=Mrecord];
	struct3 [label="hello\nworld |{ b |{c|<here> d|e}| f}| g | h"];
	struct1:f1 -> struct2:f0;
	struct1:f0 -> struct3:f1;
}

digraph dfd2{
    node[shape=record]
    subgraph level0{
        enti1 [label="Customer" shape=box];
        enti2 [label="Manager" shape=box];
    }
    subgraph cluster_level1{
        label ="Level 1";
        proc1 [label="{<f0> 1.0|<f1> One process here\n\n\n}" shape=Mrecord];
        proc2 [label="{<f0> 2.0|<f1> Other process here\n\n\n}" shape=Mrecord];
        store1 [label="<f0>    |<f1> Data store one"];
        store2 [label="<f0>   |<f1> Data store two"];
        {rank=same; store1, store2}
    }
    enti1 -> proc1
    enti2 -> proc2
    store1 -> proc1
    store2 -> proc2
    proc1 -> store2
    store2 -> proc1 
}

-->