Skip to content

Changelog

This project attempts to follow Semantic Versioning and uses Keep-a-Changelog formatting. But I make mistakes sometimes.

[4.9.0] - 2026-09-22

Added

  • StateBuilder: Viewer presentation options are keyword arguments on ViewerState and set_viewer_properties, and properties on the state: title, show_axis_lines, show_scale_bar, show_default_annotations, cross_section_background_color, projection_background_color, hide_cross_section_background_3d, and wire_frame. Unset options are left out of the state, so Neuroglancer's defaults apply.
  • StateBuilder: Camera and ViewerState.set_camera set zoom, orientation, and depth for the 2d and 3d views. Orientations are a plane name ("xy", "xz", "yz") or an [x, y, z, w] quaternion, and a Camera can be reused across many states. scale_imagery and scale_3d are shorthand for its zooms; setting them to values that conflict with a camera raises.
  • StateBuilder: SidePanel and ViewerState.set_panels(layer_list_panel=, statistics_panel=, help_panel=, selected_layer_panel=) open, close, and place Neuroglancer's side panels.
  • StateBuilder: Multi-panel layouts. Panel shows its own layers with its own view type and, optionally, its own camera, fixed or offset from the main view; row and column arrange panels, and set_layout or layout= accept them alongside the preset names.
  • StateBuilder: The tools module binds Neuroglancer tools to keys and shows them in tool palettes. Every bindable tool has a named class (tools.MeshSilhouette, tools.Alpha3d, tools.SelectSegments, tools.ShaderControlTool, ...) so editors can autocomplete them and flag typos, and each checks that it suits its layer's type. Bind them with ViewerState.add_tools(...) or on the layer, e.g. SegmentationLayer(..., tools={"H": tools.MeshSilhouette}), and group them with ToolPalette. Neuroglancer has one set of keys for the whole viewer and silently drops a tool bound to a key twice, so nglui assigns keys when the state is built: explicit keys must be free, and tools without one take the next free letter. A new Tools and Key Bindings page documents every tool.
  • StateBuilder: AnnotationLayer(active_tool="point" | "line" | "box" | "ellipsoid" | "polyline") starts an annotation layer with a drawing tool active. Neuroglancer cannot restore annotation-placing tools from key bindings -- a binding for one also drops every later binding on its layer -- so they are not offered as bindable tools.
  • StateBuilder: New layer display options. Image layers take shader_controls. Segmentation layers take segment_query, saturation, color_seed, segment_default_color, mesh_render_scale, cross_section_render_scale, hover_highlight, base_segment_coloring, ignore_null_visible_set, linked_segmentation_group, linked_segmentation_color_group, equivalences, shader_controls, and a SkeletonRendering for skeleton shaders, render modes, and line widths. Annotation layers take shader_controls and ignore_null_segment_filter.
  • StateBuilder: An extra argument on ViewerState and every layer merges raw Neuroglancer JSON over the built state last, for options nglui does not wrap, without resorting to a base_state.
  • StateBuilder: ViewerState.add_points, add_lines, add_ellipsoids, add_boxes, and add_polylines pass further AnnotationLayer options (such as active_tool or shader_controls) to the layer they create, and raise rather than silently ignore them for a layer that already exists.
  • StateBuilder: shaders.InvlerpControl declares an invlerp shader control for custom shaders.

Changed

  • StateBuilder: When several annotation layers have tags, the later layers' tag tools move to free keys instead of every layer binding Q, W, ... -- where Neuroglancer kept only one. Each tag keeps its usual key when it is free, moves are listed in a warning, and tags left without any free key are unbound, with a warning, and remain usable from the annotation panel. A single tagged layer's output is unchanged.
  • StateBuilder: A base_state's layout, zoom, and slice settings are kept unless set explicitly; layout previously always overwrote the base state's.
  • StateBuilder: Selecting a layer -- add_layer(..., selected=True), selected_layer=, or set_selected_layer -- now opens its panel; pass set_selected_layer(layer, visible=False) to select without opening. A selected layer that is not in the viewer raises.
  • StateBuilder: RGB colors are read as 0-1 when every value is within 0-1, so (1, 1, 1) is white, and as 8-bit when any value is above 1, so (128, 128, 128) is gray. Values that fit neither, such as (1.5, 0, 0), raise; they were previously clamped, which turned 8-bit colors white.
  • StateBuilder: ImageLayer.opacity and SegmentationLayer.hide_segment_zero default to None, meaning Neuroglancer's default. The previously documented defaults were never actually emitted, so output is unchanged.
  • StateBuilder: Camera, SidePanel, SkeletonRendering, Panel, ToolPalette, and the tool classes are immutable, and ViewerState.extra is read-only, so changing them after a state has been built raises instead of being silently ignored.
  • StateBuilder: shaders.InverlpControl is a deprecated alias of InvlerpControl.
  • Docs: API reference pages document modules as nglui.* rather than src.nglui.*, which changes their link anchors.

Fixed

  • StateBuilder: Several options were accepted but never written to the state: ImageLayer opacity, blend, the volume rendering options, and cross_section_render_scale; SegmentationLayer.hide_segment_zero; selected_layer; and filter_by_segmentation on local annotation layers.
  • StateBuilder: set_viewer_properties turned off infer_coordinates on every call, and set_viewer_properties(selected_layer_visible=...) raised.
  • StateBuilder: SegmentationLayer.set_view_options ignored zero values, such as not_selected_alpha=0.
  • StateBuilder: add_segments_from_data and add_segment_properties failed when given a DataMap.
  • StateBuilder: A list-valued filter_by_segmentation emitted the link dictionary instead of the list, and a string value emitted nothing.
  • StateBuilder: CoordSpaceTransform failed on a plain list for input_dimensions.
  • StateBuilder: linked_segmentation=False raised, although it is documented to mean "do not link".
  • StateBuilder: Segment ids in equivalences above 2^53 were written as bare JSON numbers, which Neuroglancer's JSON.parse rounds -- so real root ids named different segments. They are now written as strings.
  • StateBuilder: shaders.InverlpControl generated a #uicontrol directive Neuroglancer rejects.
  • Docs: Building the docs locally crashed on every API reference page because the locked pymdown-extensions was incompatible with Pygments 2.20; mkdocs build --strict now passes with no warnings.

[4.8.0] - 2026-09-11

Added

  • StateBuilder: Annotation tags can now be emitted in main Neuroglancer's format, where a tag is an annotation property of type: "bool" bound to a toggleBoolProperty hotkey, alongside the existing Spelunker format. The encoding is chosen by what the target deployment supports rather than by a version number, since Neuroglancer is not semantically versioned. Pass capabilities="main" (tracking google/neuroglancer's main branch) or capabilities="legacy" (tracking Spelunker) to ViewerState or an AnnotationLayer to choose explicitly; left unset, nglui reads the deployment's own client bundle and looks for the features in it, which needs no list of known deployments and is not fooled by a fork's version string — Spelunker reported v2.37 both before and after gaining the whole bool-property system. A deployment that cannot be identified is assumed to track main, and warns — set capabilities="legacy" or set_default_capabilities("legacy") for an unidentifiable deployment that predates bool annotation properties, which would otherwise drop the annotation layer rather than render it. That lookup only happens when a local annotation layer has tags, is cached, and never raises, so offline use is unaffected.
  • StateBuilder: The deployments nglui ships with (spelunker, google) have their capabilities declared in code, so targeting either costs no network request. declare_capabilities() records others, or corrects a shipped declaration that has gone stale without waiting for a release. A network-marked test re-validates the shipped declarations against the live deployments and is deselected by default.
  • StateBuilder: Capability lookups are cached per deployment origin for an hour and cleared with clear_capability_cache(); prefetch() warms one deliberately. Every URL pointing at the same viewer shares an entry, so a session working with many state links does not evict real answers from the cache.
  • StateBuilder: set_default_capabilities() changes the assumed encoding when the target is unknown, and NGLUI_DISABLE_CAPABILITY_PROBE=1 suppresses the lookup.
  • StateBuilder: Capabilities is a named set rather than a label: Capabilities.names() lists what nglui knows, .enabled gives the available ones, capability in capabilities tests membership, and Capabilities.from_names([...]) builds sets the "legacy"/"main" shorthands do not cover — such as a build that understands bool properties but not the tools that toggle them. Provenance is recorded in source, which is excluded from equality so a probed set compares equal to an equivalent named one.
  • StateBuilder: Tag labels become property identifiers under the modern encoding, which Neuroglancer requires to match ^[a-z][a-zA-Z0-9_]*$. Labels are sanitized the way Neuroglancer's own annotation tab sanitizes typed input, with a single warning per layer listing what changed. tag_ids sets identifiers explicitly and strict_property_ids=True turns renaming into an error.
  • Parser: annotation_property_dictionary() reports every annotation property in a layer, not only tag-like ones.
  • StateBuilder: New helpers statebuilder.helpers.make_segment_state and statebuilder.helpers.make_segment_link turn a CAVEclient plus root ids (and, optionally, segment properties) into a viewer state or a link in one line. Segment properties can be passed as a SegmentProperties object or as property JSON, and are uploaded and attached to the segmentation layer automatically.
  • StateBuilder: statebuilder.helpers.add_segment_properties_source adds an already-built segment property to a segmentation layer of an existing viewer state.
  • StateBuilder: Link-producing helpers now accept return_as="clipboard", "browser", and "viewer" in addition to "link", "url", "dict", and "json".
  • StateBuilder: statebuilder.helpers.make_neuron_neuroglancer_link gained segment_properties, segmentation_layer_name, and link_text arguments.

Fixed

  • StateBuilder: Naming a different target late, as in to_url(target_site=...), to_link, to_clipboard, to_browser and to_link_shortener, reused a viewer state that had already been built for the original target. With tag encoding now depending on the target, that would have shipped the wrong format, so the state is rebuilt when the target changes.
  • StateBuilder: Annotation layers with a cloud source silently discarded any tags set on them, and now warn.
  • StateBuilder: statebuilder.helpers.add_segment_properties_source uploaded the property JSON before checking that the target layer existed, so a mistyped layer name left an orphaned property on the state service and then failed with a bare KeyError. It now validates the layer first and names the available ones.
  • StateBuilder: Annotation tool bindings were emitted whenever bool properties were used, but the property tools landed upstream three months after the properties themselves. A deployment built in between understands the properties and not the bindings, and an unparseable tool type costs the whole layer, so bindings are now gated on that capability separately.
  • SegmentProperties: A segment with a missing value in a tag_bool_cols column was reported as carrying that tag. np.flatnonzero counts NaN as non-zero, so an absent flag read as a set one — silently, and in the direction that invents data.
  • StateBuilder: A missing value in a tag column no longer raises. tag_column with a NaN or None raised TypeError when ordering the layer's tag vocabulary, and tag_bools holding a NaN or a nullable dtype raised IndexError when used as a mask — both ordinary shapes for real data. Missing now means untagged, as do blank strings.
  • StateBuilder: Non-string tag values were silently never set. Annotations coerce their own tags to strings, but the layer's vocabulary kept the raw value, so a numeric tag_column produced a correct-looking property list with every value unset. The vocabulary is coerced to match.
  • StateBuilder: Non-ASCII tag labels are folded toward ASCII before becoming property identifiers. Accented Latin decomposes (café → cafe, previously caf), and Greek letters are replaced by their names (β-cell → beta_cell) — without which α-cell, β-cell and γ-cell all sanitized to cell and were then disambiguated into cell/cell_2/cell_3, making distinct cell types indistinguishable in output that looked correct. Scripts with no Latin reading still fall back to a generated id, with the original kept in the property description.
  • StateBuilder: A repeated entry in an annotation layer's tags produced output no viewer accepts — two properties sharing an identifier under the bool encoding, and a per-annotation value vector shorter than the property list under the Spelunker one. Repeats are now emitted once, with a warning.
  • Parser: Reading tags from a state scanned every entry of each annotation's props array, so a non-tag property whose value happened to equal 1 was reported as a set tag. Only tag-like properties are considered now.
  • StateBuilder: AnnotationLayer.tag_map returned tagTool_{i}, which is not a real tool type -- bindings use tagTool_tag{i} -- where the name and every other use in the codebase mean {tag: index}. It now returns that.
  • StateBuilder: ViewerState.add_segment_properties raised "Client must be specified" whenever a client was specified.
  • StateBuilder: statebuilder.helpers.make_neuron_neuroglancer_link failed when shortening a URL because the client was not passed through to the link shortener, and its documented shorten values of "never" and "always" did not work.

[4.7.1] - 2026-02-10

Changed

  • Bumped various dependencies.

[4.7.0] - 2026-02-10

Added

  • StateBuilder: Added Polyline annotation support, similar to other annotation types.
  • Parser: Added polyline annotation support.
  • Parser: Annotation ids are now supported as an optional return data for getting various annotation types and by default as an additional column in parser.annotation_dataframe.

[4.6.0] - 2025-11-1

Added

  • StateBuilder: Added compact argument to ViewerState.to_json_string to produce more compact JSON output by removing whitespace.

Fixed

  • StateBuilder: Fixed a bug where ViewerState.to_json_string failed if numpy types were present in the state.

[4.5.1] - 2025-08-29

Added

  • StateBuilder: Added support for passing an x,y,z column names explicitly as a list and improved docstrings around this parameter.

Fixed

  • StateBuilder: Slight improvements to docstrings.

[4.5.0] - 2025-08-15

Added

  • StateBuilder: Added RawLayer class that allows for the inclusion of layer specification directly from an existing neuroglancer state. You can also remap sources along the way using the remap_sources method, for example if you are changing states to a new endpoint. The easiest way to access is via ViewerState.add_raw_layer.
  • StateBuilder: Added ViewerState.add_segment_properties function to add segment properties within a viewer state pipeline.
  • StateBuilder: Annotation-adding functions now support adding arrays as well as dataframe columns. Data can now be None and columns can be specified completely.
  • StateBuilder: Added strip_layers function to remove layers from a viewer dictionary to ease its use as a base state.

Fixed

  • StateBuilder: The properties of a base state are better-preserved.

[4.4.2] - 2025-06-12

Added

  • StateBuilder: Added ViewerState.to_browser method that parallels ViewerState.to_url that opens a URL in the browser (h/t Ben Pedigo)

Fixed

  • StateBuilder: Fixed a bug where the to_url method did not pass the target_url argument to the link shortener.

[4.4.1] - 2025-07-01

Fixed

  • StateBuilder: Fixed segment properties to use correct auth formatting.

[4.4.0] - 2025-06-30

Added

  • StateBuilder: Added ViewerState.to_clipboard method that parallels ViewerState.to_url that copies the URL or shortened link to the clipboard (h/t Ben Pedigo)

Fixed

  • StateBuilder: In certain conditions, empty lists were incorrectly being handled.

[4.3.1] - 2025-06-25

Fixed

  • StateBuilder: More generous error handling when cloudvolume can't do its job for a given source.

[4.3.0] - 2025-06-25

  • StateBuilder: Added ViewerState.add_segments and ViewerState.add_segments_from_data to the viewerstate object. They let you add segments to a segmentation layer without breaking a pipeline. Both pass through to SegmentationLayer functions, and default to the first segmentation layer in the state if no layer name is specified.

[4.2.0] - 2025-06-25

  • StateBuilder: Added swap_visible_segments_on_move option to AnnotationLayers.

[4.1.0] - 2025-06-24

Added

  • StateBuilder: Added the pick option to layers that enable you to turn off new ID selection (pick=False).
  • StateBuilder: Added color options to add_points/add_lines/etc methods to set the layer color.
  • StateBuilder: Converted viewer_state.layers to a modified list class with names. Only objects with a .name attribute can be added to the list, and objects can be accessed by either name or index.
  • StateBuilder: Added a client argument to ViewerState that will be used as a default for any viewer_state functions that take a client.

[4.0.0]

This is a major breaking release that fully changes the way most of the library works. Code from previous verions will not work without significant changes. Parser and SegmentationProperties work as before, but the StateBuilder and EasyViewer modules have been merged with a different pattern and approach. Only the current main and Spelunker branches of Neuroglancer are supported, and the older Seung-lab branch is no longer supported. The following is a small summary of the changes.

Added

  • StateBuilder/EasyViewer: Complete rewrite joining StateBuilder and Easyviewer into a single module under the statebuilder namespace. Only supports spelunker.
  • StateBuilder: A new SkeletonManager to upload skeletons for use in Neuroglancer.
  • StateBuilder: Support for building some shaders.
  • StateBuilder: Added inference of properties from source info files$$.
  • StateBuilder: Added DataMaps to several places to allow for more flexible data handling and rendering.

Removed

  • StateBuilder/EasyViewer: Fully removed support for the Seung-lab branch of Neuroglancer. This is no longer supported and will not be updated.
  • StateBuilder/EasyViewer: Removed both main classes and their functionality. Functionality is now merged into the ViewerState class.

[3.8.2] - 2025-03-27

Fixed

  • Parser Fix issues with dimensions in the parser.

[3.8.1] - 2025-01-30

Fixed

  • StateBuilder SegmentationLayerConfig with a color column will ignore None values for data.

[3.8.0] - 2025-01-29

Added

  • StateBuilder Segmentation layers can be set to be non-interactive by default using the selectable parameter in SegmentationLayerConfig. Only works for spelunker branch.

Fixed

  • StateBuilder Adjusted sphere radius scaling to be correct.

[3.7.2] - 2024-12-18

Fixed

  • Parser: Fixed a bug in annotation parsing if some layers have tags and some do not.

[3.7.1] - 2024-12-11

Fixed

  • StateBuilder: Fixed a bug where the use of "helpers.make_state_url" (which includes most of the helpers) was not using information about urls to infer target sites.

[3.7.0] - 2024-12-10

Added

  • StateBuilder and SegmentProps: A new parameter for segment properties will create columns with a random number for you in a segment property, useful for sampling large lists of ids in neuroglancer.
  • StateBuilder: Made it so that site_utils configuration can be accessed from statebuilder.site_utils instead of needing a separate import.

Fixed

  • StateBuilder: Updated helpers to be compatible with new site_utils functionality.

[3.6.2] - 2024-11-22

Fixed

  • Fixed a bug where default configurations were overriding user configurations for target sites.

[3.6.1] - 2024-11-20

Added

  • StateBuilder: Added default configurations to make it much easier to configure links where you want them to go.

[3.5.2] - 2024-11-11

Fixed

  • Fixed behavior of statebuilder.helpers.make_neuron_neuroglancer_link to actually render the state as it used to.

[3.5.1] - 2024-11-10

Fixed

  • Removed inadvertent debug statement in the last release.
  • Fixed a bug in the expand_tags option in the annotation_dataframe function.

[3.5.0] - 2024-11-10

Added

  • StateBuilder: Added tags to Spelunker neuroglancer states! Note that this might not be available in all mainline-style deployments yet.
  • Parser: Added a number of new features to parse classic and Spelunker-style states, including layer and segment information, filtering archived layers, and a new class-level interface to make life easier.

Fixed

  • Reformatted use of target_site to be more consistent when using "spelunker" as a name.

[3.4.0] - 2024-08-26

Added

  • StateBuilder : Added use_skeleton_source as an optional argument for from_client to use a skeleton service advertised in the info service.
  • SegmentProperties New option prepend_col_name in SegmentProperties.from_dataframe will prepend the column name to the tag values.

[3.3.7] - 2024-07-24

Added

  • StateBuilder New option on StateBuilder.render_as: return_as="short" will upload the state and return a short URL to the state.
  • StateBuilder New helper statebuilder.helpers.segment_property_link will quickly generate a basic link from a segment property.

Changed

  • Default "spelunker" neuroglancer deployment was changed.

Fixed

  • StateBuilder Fixed a bug when setting the background color in Spelunker sites.
  • StateBuilder Fixed a bug in using segment property maps.

[3.3.6] - 2024-07-23

Added

  • SegmentProperties New method "label_format_map" allows you to build arbitrary label formats from dataframe columns using

Fixed

  • SegmentProperties Nulls in label columns are ignored as intended.

[3.3.5] - 2024-07-19

Added

  • SegmentProperties The label_col argument in SegmentProperties.from_dataframe can now take a list of column names. Labels are concatenated with a seperator set with label_separator that defaults to an underscore (_).
  • StateBuilder Recent segment property arguments included in the segment property maps for SegmentationLayerConfigs.

Changed

  • SegmentProperties All property names (the id field) are coerced to strings, as required.

[3.3.4] - 2024-07-19

Fixed

  • SegmentProperties: Fixed a bug in building tags from dataframes.

[3.3.3] - 2024-07-19

Added

  • SegmentProperties: Tag columns will now automatically disambuguate tags if they are duplicated in different columns of the dataframe. For example, if you had "my_column" with value "type_a" and "their_column" also with value "type_a", the tag would become "my_column:type_a" and "their_column:type_a". This can be turned off by setting the allow_disambiguation argument to False.
  • SegmentProperties: The SegmentProperties.from_dataframe method now has a allow_disambiguation argument to control whether the disambiguation above is performed.

Changed

  • SegmentProperties: Tag property generation performance for long dataframes is improved.

Fixed

  • SegmentProperties: Fixed a bug in handling tag columns with categorical dtype.

[3.3.2] - 2024-07-18

Changed

  • StateBuilder: When providing a URL but not a client, you will now get a warning that target-site cannot be inferred.

Fixed

  • Small changes to work in python 3.8, because the only issue was type hints.

[3.3.1] - 2024-07-12

Fixed

  • Removed debug statements left in after testing.

[3.3.0] - 2024-07-12

Added

  • StateBuilder : Added a skeleton_source and skeleton_shader parameter to SegmentationLayerConfig to put skeleton info into the state.
  • EasyViewer : Added set_skeleton_source and set_skeleton_shader methods to EasyViewer.

Fixed

  • SegmentProperties : Better handling of null values (nan, None, and empty strings) in tag columns, all of which are ignored.
  • Various requirements updates that were necessary.

[3.2.1] - 2024-07-11

Added

  • SegmentProperties : Added a .to_dataframe() method for segment properties to convert them back to a dataframe.

[3.2.0] - 2024-07-10

Added

  • SegmentProperties : A new module has been created to build Segment Properties, which are a method of organizing information about segment IDs in a Neuroglancer segmentation layer. Segment Properties only work with the newer Google/Spelunker branches, and offer a way to add browsable and searchable metadata to segments that can be viewed in the viewer.
  • StateBuilder : Various changes were made to support segment properties from both explicit URLs and from data-driven mapping. See documentation for details.

Changed

  • StateBuilder : Image layers for spelunker now use native contrast controls.

Fixed

  • StateBuilder : Improvements to the use of target_site and url_prefix in general when both creating statebuilder and rendering states. In particular, the values in render_state should now correctly take precendence over the values in StateBuilder when both are set.

[3.1.0] - 2024-05-29

Added

  • The Parser function annotation_dataframe now has a parameter expand_tags that will create a boolean column for each tag in each annotation layer indicating if the annotation has that tag or not.

Fixed

  • Parser : Fixed bugs in getting bounding box annotations and in getting linked segmentations for Spelunker-style states.

[3.0.3] – 2024-4-30

Fixed

  • StateBuilder : Fixed a bug preventing tags from being set.

[3.0.2] – 2024-1-19

Fixed

  • Improve consistency of target_site argument in StateBuilder.

[3.0.1] – 2023-12-20

Fixed

  • Fixed an unnecessary import that caused an error.
  • Added python package neuroglancer to the requirements so versions with errors or incompatibilities can be avoided.

[3.0.0] – 2023-12-20

Added

  • StateBuilder and EasyViewer Substantial refactor, adding basic support for the main Google Neuroglancer branch and the Cave-Explorer/Spelunker deployments coming online with CAVE. Most functionality should be seamless, with the client checking for version. Currently, only the Cave-Explorer and Spelunker deployments are fully supported for this automated version check, as the response depends on a build-specific version.json file. If no automated version is set or one wants to override any settings, the target_site argument can be set to cave-explorer or mainline for the main Google branch or seunglab for the older Seung-lab branch.

[2.14.1] - 2023-05-14

Fixed

  • StateBuilder Recast all layer names as strings to avoid issues with numeric.

[2.14.0] - 2023-04-12

Fixed

  • StateBuilder Fixed various issues with stacking multiple points into a single row.

[2.13.0] - 2023-02-21

Added

  • StateBuilder Adding a helper function for basic line annotation states
  • StateBuilder Helper functions can take split_position argument.

[2.12.1] - 2023-01-30

Fixed

  • StateBuilder Handle None data correctly with mapping sets.

[2.12.0] - 2023-01-30

Added

  • EasyViewer At long last, you can set the background color in set_view_options with the argument background_color. As elsewhere, this uses webcolors to deal with color parsing, so it can take names or rgb or hex.
  • StateBuilder StateBuilder can now set the background color in in view_kws with the key background_color.
  • StateBuilder Mapping sets offer a new and simple iapproach to having multiple dataframes with different mapping rules.
  • StateBuilder Mapping rules can handle dataframes where the point position components are split across different columns, assuming that the suffices are always _x, _y, and _z. Set split_positions=True in the mapping rule to use this.

Changed

  • StateBuilder Mapping rules now default to set_position = True.

[2.10.0] - 2022-08-11

Added

  • StateBuilder A number of helper functions to produce common states have been added to statebuilder. You can use the function make_neuron_neuroglancer_link to generate a link with one or more root ids and, optionally, their synapses. Further, you can use the function make_synapse_neuroglancer_link to generate a state from a synapse dataframe.
  • StateBuilder Linked segmentations and annotation groups now support multiple columns. For linked segmentations, this adds multiple root ids per annotation. For groups, each unique combinations of values gets its own group.
  • StateBuilder and EasyViewer Colors for layers or segment ids can now be in RGB tuples with values between 0--1 or CSS3 named colors, not just hex.
  • StateBuilder When using the client= argument, viewer resolution is inferred from the client info and does not need to be passed as an additional argument.

Fixed

  • StateBuilder For SegmentationLayerConfigs, adding selection maps via add_selection_map is more robust.

[2.7.2] - 2021-06-13

Fixed

  • StateBuilder Multipoint now works with set_position.

[2.7.1] - 2021-06-13

Fixed

  • StateBuilder Multipoint performance has improved.

[2.7.0] - 2021-06-13

Added

  • StateBuilder In all Mappers, setting multipoint=True will allow the point columns to contain multiple points per row, with the other columns assigned to all points in the row. Note that for Mappers with multiple point columns (e.g. LineMappers, SphereMappers, and BoundingBoxMappers), the number of values must be the same in both columns.

[2.6.0] - 2021-06-11

Added

  • StateBuilder You can now set a timestamp in SegmentationLayerConfig. Either datetime or unix epoch are allowed.

[2.5.0]

  • StateBuilder You can now set split points using a SplitPointMapper

[2.4.0] - 2021-01-22

Added

  • Parser: get_selected_ids does what it says for a layer.

Fixed

  • Parser: Getting annotations with tags/descriptions/etc works.
  • StateBuilder: return_as parameter works with ChainedStateBuilder now

[2.3.1] - 2021-01-15

Fixed

  • Remove pytables from requirements, because the pip install does not work on OS X (at least). If you want to run tests, you will need to install either with pip install tables if you're on a system where that works or conda install pytables otherwise.

[2.3.0] - 2021-01-14

Added

  • StateBuilder: At long last, a BoundingBoxMapper
  • StateBuilder: A from_client function to generate simple image & segmentation states from a FrameworkClient instance.
  • StateBuilder: Statebuilder can take a client to configure certain default parameters.
  • StateBuilder: SegmentationLayerConfig explicitly takes some view keyword arguments: alpha_selected, alpha_3d, and alpha_unselected.
  • Parser: New options to get annotation groups and group ids for all annotations.
  • Parser: New function to extract multicut information from a state.

Fixed

  • StateBuilder: Fixed a scenario where int64s get altered by a conversion through floats.

Changed

  • Statebuilder: Behind the scenes refactoring that should not affect use.

[2.2.1] — 2020-10-21

Fixed

  • StateBuilder: Setting view options for Graphene segmentation layers now works
  • Parser: Now imported as a property of nglui if you import as import nglui.
  • Small bug fixes.

[2.2.0] — 2020-09-04

Added

  • Parser submodule nglui.parser for quickly extracting data from neuroglancer states. This should remove some of the boilerplate one writes every time you want to get data out of a state.

  • AnnotationLayerConfig now can take arguments about user interactions such as filtering by segmentation and bracket shortcuts showing segmentations.

Changed

  • Default behavior for annotation layers now has filtering by segmentation turned off and bracket shortcuts showing segmentations turned on.

[2.1.1] — 2020-08-17

Changed

  • Switched from numpy isnan to pandas isnull in StateBuilder. This allows nullable pandas Int64 dtypes to work as columns

[2.1.0] — 2020-07-14

Added

  • Grouped annotations in nglite base and EasyViewer's annotation module. This feature lets you create annotation groups in the Seung-lab branches of Neuroglancer, where multiple annotations are related to one another. It works by passing a list of already-created annotations and getting a new "CollectionAnnotation" that groups them together.

  • Grouped annotations in StateBuilder. PointMapper, SphereMapper, and LineMapper objects can take a 'group_column' value. Data in this column is intended to be numeric or NaN, and rows that share the same value are grouped together. At the moment, this feature only works within individual mapper objects.

  • Two options for linked segmentation ids on grouped annotations: gather_linked_segmentations assigns all linked ids of objects within the group to the group annotation itself (True by default). share_linked_segmentations will add all linked objects within the group to all annotations within the group (False by default).

  • array_data option on AnnotationLayerConfig for simple cases where you just want to map Nx3 arrays to points, a pair of Nx3 arrays to lines, or an Nx3 + N array to spheres (centers+radii).

Fixed

  • Bug in GL Shader that caused StateBuilder to fail when constrast_controls was set to True.

[2.0.2] - 2020-06-18

Added

  • Custom GL Shaders (see documentation) for image and segmentation layers can be set in EasyViewer.

  • JSON State Server can be set in EasyViewer.

  • Added contrast_controls argument to statebuilder.ImageLayerConfig to provide brightness/contrast controls to layer through GL Shader.

  • Added state_server argument to statebuilder.StateBuilder to pre-set state server endpoints.

[2.0.1] - 2020-04-27

Fixed

  • Removed unnecessary parts of the included neuroglancer module (nglite) for cleanliness.

[2.0.0] - 2020-04-25

This update significantly changed the underlying nature and goals of the Neuroglancer Annotation UI project.

Added

  • StateBuilder module for rule-based method of generating Neuroglancer states from Pandas DataFrames.

  • Added various features to make use of specific aspects of the Seung-lab Neuroglancer fork.

Changed

  • Removed dependency on neuroglancer python package, which was not designed for the Seung-Lab Neuroglancer fork and had some difficult install requirements for features that were unnecessary for this project demanded.

Removed

  • Dynamic state management via python server.