Version v1.9.0 Jan 30, 2026

1. Format Overview

1.1. Namespace – HDMF Common

• Description: Common data structures provided by HDMF

• Name: hdmf-common

• Full Name: HDMF Common

• Version: 1.9.0

• Authors:

  • Andrew Tritt

  • Oliver Ruebel

  • Ryan Ly

  • Ben Dichter

• Contacts:

  • ajtritt@lbl.gov

  • oruebel@lbl.gov

  • rly@lbl.gov

  • bdichter@lbl.gov

• Schema:

  • doc: base data types

  • source: base.yaml

  • title: Base data types

  • doc: data types for a column-based table

  • source: table.yaml

  • title: Table data types

  • doc: data types for different types of sparse matrices

  • source: sparse.yaml

  • title: Sparse data types

  • doc: data types for storing references to web accessible resources

  • source: resources.yaml

  • title: Resource reference data types

1.2. Type Hierarchy

• Container

  • CSRMatrix

  • DynamicTable

    • AlignedDynamicTable

    • MeaningsTable

  • HERD

  • SimpleMultiContainer

• Data

  • ElementIdentifiers

  • VectorData

    • DynamicTableRegion

    • VectorIndex

2. Type Specifications

2.1. Base data types

base data types

2.1.1. Data

Overview: An abstract data type for a dataset.

• Primitive Type: Dataset

• Subtypes: ElementIdentifiers, VectorIndex, DynamicTableRegion, VectorData

• Source filename: base.yaml

• Source Specification: see Section 3.2.1

2.1.2. Container

Overview: An abstract data type for a group storing collections of data and metadata. Base type for all data and metadata containers.

• Primitive Type: Group

• Subtypes: AlignedDynamicTable, HERD, DynamicTable, CSRMatrix, MeaningsTable, SimpleMultiContainer

• Source filename: base.yaml

• Source Specification: see Section 3.2.2

2.1.3. SimpleMultiContainer

Overview: A simple Container for holding onto multiple containers.

SimpleMultiContainer extends Container and includes all elements of Container with the following additions or changes.

• Extends: Container

• Primitive Type: Group

• Inherits from: Container

• Source filename: base.yaml

• Source Specification: see Section 3.2.3

Table 2.1 Datasets, Links, and Attributes contained in <SimpleMultiContainer>

Id	Type	Description
<SimpleMultiContainer>	Group	Top level Group for <SimpleMultiContainer> Neurodata Type: SimpleMultiContainer Extends: Container
.<Data>	Dataset	Data objects held within this SimpleMultiContainer. Extends: Data Quantity: 0 or more

Table 2.2 Groups contained in <SimpleMultiContainer>

Id	Type	Description
<SimpleMultiContainer>	Group	Top level Group for <SimpleMultiContainer> Neurodata Type: SimpleMultiContainer Extends: Container
.<Container>	Group	Container objects held within this SimpleMultiContainer. Extends: Container Quantity: 0 or more

2.1.3.1. Groups: <Container>

Container objects held within this SimpleMultiContainer.

• Extends: Container

• Quantity: 0 or more

2.2. Table data types

data types for a column-based table

2.2.1. VectorData

Overview: An n-dimensional dataset representing a column of a DynamicTable. If used without an accompanying VectorIndex, first dimension is along the rows of the DynamicTable and each step along the first dimension is a cell of the larger table. VectorData can also be used to represent a ragged array if paired with a VectorIndex. This allows for storing arrays of varying length in a single cell of the DynamicTable by indexing into this VectorData. The first vector is at VectorData[0:VectorIndex[0]]. The second vector is at VectorData[VectorIndex[0]:VectorIndex[1]], and so on.

VectorData extends Data and includes all elements of Data with the following additions or changes.

• Extends: Data

• Primitive Type: Dataset

• Dimensions: [[‘dim0’], [‘dim0’, ‘dim1’], [‘dim0’, ‘dim1’, ‘dim2’], [‘dim0’, ‘dim1’, ‘dim2’, ‘dim3’]]

• Shape: [[None], [None, None], [None, None, None], [None, None, None, None]]

• Inherits from: Data

• Subtypes: VectorIndex, DynamicTableRegion

• Source filename: table.yaml

• Source Specification: see Section 3.3.1

Table 2.3 Datasets, Links, and Attributes contained in <VectorData>

Id	Type	Description
<VectorData>	Dataset	Top level Dataset for <VectorData> Neurodata Type: VectorData Extends: Data Dimensions: [[‘dim0’], [‘dim0’, ‘dim1’], [‘dim0’, ‘dim1’, ‘dim2’], [‘dim0’, ‘dim1’, ‘dim2’, ‘dim3’]] Shape: [[None], [None, None], [None, None, None], [None, None, None, None]]
.description	Attribute	Description of what these vectors represent. Data Type: text Name: description

2.2.2. VectorIndex

Overview: Used with VectorData to encode a ragged array. An array of indices into the first dimension of the target VectorData, forming a map between the rows of a DynamicTable and the indices of the VectorData. The name of the VectorIndex is expected to be the name of the target VectorData object followed by “_index”.

VectorIndex extends VectorData and includes all elements of VectorData with the following additions or changes.

• Extends: VectorData

• Primitive Type: Dataset

• Data Type: uint8

• Dimensions: [‘num_rows’]

• Shape: [None]

• Inherits from: VectorData, Data

• Source filename: table.yaml

• Source Specification: see Section 3.3.2

Table 2.4 Datasets, Links, and Attributes contained in <VectorIndex>

Id	Type	Description
<VectorIndex>	Dataset	Top level Dataset for <VectorIndex> Neurodata Type: VectorIndex Extends: VectorData Data Type: uint8 Dimensions: [‘num_rows’] Shape: [None]
.target	Attribute	Reference to the target dataset that this index applies to. Data Type: object reference to VectorData Name: target

2.2.3. ElementIdentifiers

Overview: A list of unique identifiers for values within a dataset, e.g. rows of a DynamicTable.

ElementIdentifiers extends Data and includes all elements of Data with the following additions or changes.

• Extends: Data

• Primitive Type: Dataset

• Data Type: int32

• Dimensions: [‘num_elements’]

• Shape: [None]

• Default Name: element_id

• Inherits from: Data

• Source filename: table.yaml

• Source Specification: see Section 3.3.3

2.2.4. DynamicTableRegion

Overview: DynamicTableRegion provides a link from one table to an index or region of another. The table attribute is a link to another DynamicTable, indicating which table is referenced, and the data is int(s) indicating the row(s) (0-indexed) of the target array. DynamicTableRegion`s can be used to associate rows with repeated meta-data without data duplication. They can also be used to create hierarchical relationships between multiple `DynamicTable`s. `DynamicTableRegion objects may be paired with a VectorIndex object to create ragged references, so a single cell of a DynamicTable can reference many rows of another DynamicTable.

DynamicTableRegion extends VectorData and includes all elements of VectorData with the following additions or changes.

• Extends: VectorData

• Primitive Type: Dataset

• Data Type: int32

• Dimensions: [‘num_rows’]

• Shape: [None]

• Inherits from: VectorData, Data

• Source filename: table.yaml

• Source Specification: see Section 3.3.4

Table 2.5 Datasets, Links, and Attributes contained in <DynamicTableRegion>

Id	Type	Description
<DynamicTableRegion>	Dataset	Top level Dataset for <DynamicTableRegion> Neurodata Type: DynamicTableRegion Extends: VectorData Data Type: int32 Dimensions: [‘num_rows’] Shape: [None]
.table	Attribute	Reference to the DynamicTable object that this region applies to. Data Type: object reference to DynamicTable Name: table
.description	Attribute	Description of what this table region points to. Data Type: text Name: description

2.2.5. DynamicTable

Overview: A group containing multiple datasets that are aligned on the first dimension (Currently, this requirement is left up to APIs to check and enforce). These datasets represent different columns in the table. Apart from a column that contains unique identifiers for each row, there are no other required datasets. Users are free to add any number of custom VectorData objects (columns) here. DynamicTable also supports ragged array columns, where each element can be of a different size. To add a ragged array column, use a VectorIndex type to index the corresponding VectorData type. See documentation for VectorData and VectorIndex for more details. Unlike a compound data type, which is analogous to storing an array-of-structs, a DynamicTable can be thought of as a struct-of-arrays. This provides an alternative structure to choose from when optimizing storage for anticipated access patterns. Additionally, this type provides a way of creating a table without having to define a compound type up front. Although this convenience may be attractive, users should think carefully about how data will be accessed. DynamicTable is more appropriate for column-centric access, whereas a dataset with a compound type would be more appropriate for row-centric access. Finally, data size should also be taken into account. For small tables, performance loss may be an acceptable trade-off for the flexibility of a DynamicTable.

DynamicTable extends Container and includes all elements of Container with the following additions or changes.

• Extends: Container

• Primitive Type: Group

• Inherits from: Container

• Subtypes: MeaningsTable, AlignedDynamicTable

• Source filename: table.yaml

• Source Specification: see Section 3.3.5

Table 2.6 Datasets, Links, and Attributes contained in <DynamicTable>

Id	Type	Description
<DynamicTable>	Group	Top level Group for <DynamicTable> Neurodata Type: DynamicTable Extends: Container
.colnames	Attribute	The names of the columns in this table. This should be used to specify an order to the columns. Data Type: text Dimensions: [‘num_columns’] Shape: [None] Name: colnames
.description	Attribute	Description of what is in this dynamic table. Data Type: text Name: description
.id	Dataset	Array of unique identifiers for the rows of this dynamic table. Extends: ElementIdentifiers Dimensions: [‘num_rows’] Shape: [None] Name: id
.<VectorData>	Dataset	Vector columns, including index columns, of this dynamic table. Extends: VectorData Quantity: 0 or more

Table 2.7 Groups contained in <DynamicTable>

Id	Type	Description
<DynamicTable>	Group	Top level Group for <DynamicTable> Neurodata Type: DynamicTable Extends: Container
.meanings_tables	Group	Group containing MeaningsTable objects that provide meanings for values in VectorData columns within this DynamicTable. Quantity: 0 or 1 Name: meanings_tables

2.2.5.1. Groups: meanings_tables

Group containing MeaningsTable objects that provide meanings for values in VectorData columns within this DynamicTable.

• Quantity: 0 or 1

• Name: meanings_tables

Table 2.8 Groups contained in <meanings_tables>

Id	Type	Description
meanings_tables	Group	Top level Group for meanings_tables Quantity: 0 or 1 Name: meanings_tables
.<MeaningsTable>	Group	MeaningsTable objects that provide meanings for values in VectorData columns within this DynamicTable. Tables should be named according to the column they provide meanings for with a “_meanings” suffix. e.g., if a VectorData column is named “stimulus_type”, the corresponding MeaningsTable should be named “stimulus_type_meanings”. Extends: MeaningsTable Quantity: 0 or more

2.2.5.2. Groups: meanings_tables/<MeaningsTable>

MeaningsTable objects that provide meanings for values in VectorData columns within this DynamicTable. Tables should be named according to the column they provide meanings for with a “_meanings” suffix. e.g., if a VectorData column is named “stimulus_type”, the corresponding MeaningsTable should be named “stimulus_type_meanings”.

• Extends: MeaningsTable

• Quantity: 0 or more

2.2.6. AlignedDynamicTable

Overview: DynamicTable container that supports storing a collection of sub-tables. Each sub-table is a DynamicTable itself that is aligned with the main table by row index. I.e., all DynamicTables stored in this group MUST have the same number of rows. This type effectively defines a 2-level table in which the main data is stored in the main table implemented by this type and additional columns of the table are grouped into categories, with each category being represented by a separate DynamicTable stored within the group.

AlignedDynamicTable extends DynamicTable and includes all elements of DynamicTable with the following additions or changes.

• Extends: DynamicTable

• Primitive Type: Group

• Inherits from: DynamicTable, Container

• Source filename: table.yaml

• Source Specification: see Section 3.3.6

Table 2.9 Datasets, Links, and Attributes contained in <AlignedDynamicTable>

Id	Type	Description
<AlignedDynamicTable>	Group	Top level Group for <AlignedDynamicTable> Neurodata Type: AlignedDynamicTable Extends: DynamicTable
.categories	Attribute	The names of the categories in this AlignedDynamicTable. Each category is represented by one DynamicTable stored in the parent group. This attribute should be used to specify an order of categories and the category names must match the names of the corresponding DynamicTable in the group. Data Type: text Dimensions: [‘num_categories’] Shape: [None] Name: categories

Table 2.10 Groups contained in <AlignedDynamicTable>

Id	Type	Description
<AlignedDynamicTable>	Group	Top level Group for <AlignedDynamicTable> Neurodata Type: AlignedDynamicTable Extends: DynamicTable
.<DynamicTable>	Group	A DynamicTable representing a particular category for columns in the AlignedDynamicTable parent container. The table MUST be aligned with (i.e., have the same number of rows) as all other DynamicTables stored in the AlignedDynamicTable parent container. The name of the category is given by the name of the DynamicTable and its description by the description attribute of the DynamicTable. Extends: DynamicTable Quantity: 0 or more

2.2.6.1. Groups: <DynamicTable>

A DynamicTable representing a particular category for columns in the AlignedDynamicTable parent container. The table MUST be aligned with (i.e., have the same number of rows) as all other DynamicTables stored in the AlignedDynamicTable parent container. The name of the category is given by the name of the DynamicTable and its description by the description attribute of the DynamicTable.

• Extends: DynamicTable

• Quantity: 0 or more

2.2.7. MeaningsTable

Overview: A table to store information about the meanings of values in a linked VectorData object. All possible values of the linked VectorData object should be present in the ‘value’ column of this table, even if the value is not observed in the data. Additional columns may be added to store additional metadata about each value. The name of the MeaningsTable should correspond to the name of the linked VectorData object with a “_meanings” suffix. e.g., if the linked VectorData object is named “stimulus_type”, the corresponding MeaningsTable should be named “stimulus_type_meanings”.

MeaningsTable extends DynamicTable and includes all elements of DynamicTable with the following additions or changes.

• Extends: DynamicTable

• Primitive Type: Group

• Inherits from: DynamicTable, Container

• Source filename: table.yaml

• Source Specification: see Section 3.3.7

Table 2.11 Datasets, Links, and Attributes contained in <MeaningsTable>

Id	Type	Description
<MeaningsTable>	Group	Top level Group for <MeaningsTable> Neurodata Type: MeaningsTable Extends: DynamicTable
.value	Dataset	The value of a row in the linked VectorData object. Extends: VectorData Name: value
.meaning	Dataset	The meaning of the value in the linked VectorData object. Extends: VectorData Data Type: text Name: meaning
.target	Link	Link to the VectorData object for which this table provides meanings. Target Type VectorData Name: target

Table 2.12 Groups contained in <MeaningsTable>

Id	Type	Description
<MeaningsTable>	Group	Top level Group for <MeaningsTable> Neurodata Type: MeaningsTable Extends: DynamicTable
.target	Link	Link to the VectorData object for which this table provides meanings. Target Type VectorData Name: target

2.3. Sparse data types

data types for different types of sparse matrices

2.3.1. CSRMatrix

Overview: A compressed sparse row matrix. Data are stored in the standard CSR format, where column indices for row i are stored in indices[indptr[i]:indptr[i+1]] and their corresponding values are stored in data[indptr[i]:indptr[i+1]].

CSRMatrix extends Container and includes all elements of Container with the following additions or changes.

• Extends: Container

• Primitive Type: Group

• Inherits from: Container

• Source filename: sparse.yaml

• Source Specification: see Section 3.4.1

Table 2.13 Datasets, Links, and Attributes contained in <CSRMatrix>

Id	Type	Description
<CSRMatrix>	Group	Top level Group for <CSRMatrix> Neurodata Type: CSRMatrix Extends: Container
.shape	Attribute	The shape (number of rows, number of columns) of this sparse matrix. Data Type: uint Dimensions: [‘number of rows, number of columns’] Shape: [2] Name: shape
.indices	Dataset	The column indices. Data Type: uint Dimensions: [‘number of non-zero values’] Shape: [None] Name: indices
.indptr	Dataset	The row index pointer. Data Type: uint Dimensions: [‘number of rows in the matrix + 1’] Shape: [None] Name: indptr
.data	Dataset	The non-zero values in the matrix. Dimensions: [‘number of non-zero values’] Shape: [None] Name: data

2.4. Resource reference data types

data types for storing references to web accessible resources

2.4.1. HERD

Overview: HDMF External Resources Data Structure. A set of six tables for tracking external resource references in a file or across multiple files.

HERD extends Container and includes all elements of Container with the following additions or changes.

• Extends: Container

• Primitive Type: Group

• Inherits from: Container

• Source filename: resources.yaml

• Source Specification: see Section 3.5.1

Table 2.14 Datasets, Links, and Attributes contained in <HERD>

Id	Type	Description
<HERD>	Group	Top level Group for <HERD> Neurodata Type: HERD Extends: Container
.keys	Dataset	A table for storing user terms that are used to refer to external resources. Extends: Data Data Type: Compound data type with the following elements: key: The user term that maps to one or more resources in the resources table, e.g., “human”. (dtype= text ) Dimensions: [‘num_rows’] Shape: [None] Name: keys
.files	Dataset	A table for storing object ids of files used in external resources. Extends: Data Data Type: Compound data type with the following elements: file_object_id: The object id (UUID) of a file that contains objects that refers to external resources. (dtype= text ) Dimensions: [‘num_rows’] Shape: [None] Name: files
.entities	Dataset	A table for mapping user terms (i.e., keys) to resource entities. Extends: Data Data Type: Compound data type with the following elements: entity_id: The compact uniform resource identifier (CURIE) of the entity, in the form [prefix]:[unique local identifier], e.g., ‘NCBI_TAXON:9606’. (dtype= text ) entity_uri: The URI for the entity this reference applies to. This can be an empty string. e.g., https://www.ncbi.nlm.nih.gov/Taxonomy/Browser/wwwtax.cgi?mode=info&id=9606 (dtype= text ) Dimensions: [‘num_rows’] Shape: [None] Name: entities
.objects	Dataset	A table for identifying which objects in a file contain references to external resources. Extends: Data Data Type: Compound data type with the following elements: files_idx: The row index to the file in the files table containing the object. (dtype= uint ) object_id: The object id (UUID) of the object. (dtype= text ) object_type: The data type of the object. (dtype= text ) relative_path: The relative path from the data object with the object_id to the dataset or attribute with the value(s) that is associated with an external resource. This can be an empty string if the object is a dataset that contains the value(s) that is associated with an external resource. (dtype= text ) field: The field within the compound data type using an external resource. This is used only if the dataset or attribute is a compound data type; otherwise this should be an empty string. (dtype= text ) Dimensions: [‘num_rows’] Shape: [None] Name: objects
.object_keys	Dataset	A table for identifying which objects use which keys. Extends: Data Data Type: Compound data type with the following elements: objects_idx: The row index to the object in the objects table that holds the key (dtype= uint ) keys_idx: The row index to the key in the keys table. (dtype= uint ) Dimensions: [‘num_rows’] Shape: [None] Name: object_keys
.entity_keys	Dataset	A table for identifying which keys use which entity. Extends: Data Data Type: Compound data type with the following elements: entities_idx: The row index to the entity in the entities table. (dtype= uint ) keys_idx: The row index to the key in the keys table. (dtype= uint ) Dimensions: [‘num_rows’] Shape: [None] Name: entity_keys

3. Schema Sources

Source Specification: see Section 3.1

3.1. Namespace – HDMF Common

Description: see Section 1.1

YAML Specification:

name: hdmf-common
doc: Common data structures provided by HDMF
schema:
- doc: base data types
  source: base.yaml
  title: Base data types
- doc: data types for a column-based table
  source: table.yaml
  title: Table data types
- doc: data types for different types of sparse matrices
  source: sparse.yaml
  title: Sparse data types
- doc: data types for storing references to web accessible resources
  source: resources.yaml
  title: Resource reference data types
full_name: HDMF Common
version: 1.9.0
author:
- Andrew Tritt
- Oliver Ruebel
- Ryan Ly
- Ben Dichter
contact:
- ajtritt@lbl.gov
- oruebel@lbl.gov
- rly@lbl.gov
- bdichter@lbl.gov

3.2. Base data types

base data types

3.2.1. Data

Description: see Section 2.1.1

YAML Specification:

data_type_def: Data
doc: An abstract data type for a dataset.

3.2.2. Container

Description: see Section 2.1.2

YAML Specification:

data_type_def: Container
doc: An abstract data type for a group storing collections of data and metadata.
  Base type for all data and metadata containers.

3.2.3. SimpleMultiContainer

Extends: Container

Description: see Section 2.1.3

YAML Specification:

data_type_def: SimpleMultiContainer
data_type_inc: Container
doc: A simple Container for holding onto multiple containers.
datasets:
- data_type_inc: Data
  doc: Data objects held within this SimpleMultiContainer.
  quantity: '*'
groups:
- data_type_inc: Container
  doc: Container objects held within this SimpleMultiContainer.
  quantity: '*'

3.3. Table data types

data types for a column-based table

3.3.1. VectorData

Extends: Data

Description: see Section 2.2.1

YAML Specification:

data_type_def: VectorData
data_type_inc: Data
dims:
- - dim0
- - dim0
  - dim1
- - dim0
  - dim1
  - dim2
- - dim0
  - dim1
  - dim2
  - dim3
shape:
- - null
- - null
  - null
- - null
  - null
  - null
- - null
  - null
  - null
  - null
doc: An n-dimensional dataset representing a column of a DynamicTable. If used
  without an accompanying VectorIndex, first dimension is along the rows of the
  DynamicTable and each step along the first dimension is a cell of the larger
  table. VectorData can also be used to represent a ragged array if paired with
  a VectorIndex. This allows for storing arrays of varying length in a single
  cell of the DynamicTable by indexing into this VectorData. The first vector is
  at VectorData[0:VectorIndex[0]]. The second vector is at
  VectorData[VectorIndex[0]:VectorIndex[1]], and so on.
attributes:
- name: description
  dtype: text
  doc: Description of what these vectors represent.

3.3.2. VectorIndex

Extends: VectorData

Description: see Section 2.2.2

YAML Specification:

data_type_def: VectorIndex
data_type_inc: VectorData
dtype: uint8
dims:
- num_rows
shape:
- null
doc: Used with VectorData to encode a ragged array. An array of indices into the
  first dimension of the target VectorData, forming a map between the rows of a
  DynamicTable and the indices of the VectorData. The name of the VectorIndex is
  expected to be the name of the target VectorData object followed by "_index".
attributes:
- name: target
  dtype:
    target_type: VectorData
    reftype: object
  doc: Reference to the target dataset that this index applies to.

3.3.3. ElementIdentifiers

Extends: Data

Description: see Section 2.2.3

YAML Specification:

data_type_def: ElementIdentifiers
data_type_inc: Data
default_name: element_id
dtype: int32
dims:
- num_elements
shape:
- null
doc: A list of unique identifiers for values within a dataset, e.g. rows of a
  DynamicTable.

3.3.4. DynamicTableRegion

Extends: VectorData

Description: see Section 2.2.4

YAML Specification:

data_type_def: DynamicTableRegion
data_type_inc: VectorData
dtype: int32
dims:
- num_rows
shape:
- null
doc: DynamicTableRegion provides a link from one table to an index or region of
  another. The `table` attribute is a link to another `DynamicTable`, indicating
  which table is referenced, and the data is int(s) indicating the row(s)
  (0-indexed) of the target array. `DynamicTableRegion`s can be used to
  associate rows with repeated meta-data without data duplication. They can also
  be used to create hierarchical relationships between multiple `DynamicTable`s.
  `DynamicTableRegion` objects may be paired with a `VectorIndex` object to
  create ragged references, so a single cell of a `DynamicTable` can reference
  many rows of another `DynamicTable`.
attributes:
- name: table
  dtype:
    target_type: DynamicTable
    reftype: object
  doc: Reference to the DynamicTable object that this region applies to.
- name: description
  dtype: text
  doc: Description of what this table region points to.

3.3.5. DynamicTable

Extends: Container

Description: see Section 2.2.5

YAML Specification:

data_type_def: DynamicTable
data_type_inc: Container
doc: A group containing multiple datasets that are aligned on the first
  dimension (Currently, this requirement is left up to APIs to check and
  enforce). These datasets represent different columns in the table. Apart from
  a column that contains unique identifiers for each row, there are no other
  required datasets. Users are free to add any number of custom VectorData
  objects (columns) here. DynamicTable also supports ragged array columns, where
  each element can be of a different size. To add a ragged array column, use a
  VectorIndex type to index the corresponding VectorData type. See documentation
  for VectorData and VectorIndex for more details. Unlike a compound data type,
  which is analogous to storing an array-of-structs, a DynamicTable can be
  thought of as a struct-of-arrays. This provides an alternative structure to
  choose from when optimizing storage for anticipated access patterns.
  Additionally, this type provides a way of creating a table without having to
  define a compound type up front. Although this convenience may be attractive,
  users should think carefully about how data will be accessed. DynamicTable is
  more appropriate for column-centric access, whereas a dataset with a compound
  type would be more appropriate for row-centric access. Finally, data size
  should also be taken into account. For small tables, performance loss may be
  an acceptable trade-off for the flexibility of a DynamicTable.
attributes:
- name: colnames
  dtype: text
  dims:
  - num_columns
  shape:
  - null
  doc: The names of the columns in this table. This should be used to specify an
    order to the columns.
- name: description
  dtype: text
  doc: Description of what is in this dynamic table.
datasets:
- name: id
  data_type_inc: ElementIdentifiers
  dims:
  - num_rows
  shape:
  - null
  doc: Array of unique identifiers for the rows of this dynamic table.
- data_type_inc: VectorData
  doc: Vector columns, including index columns, of this dynamic table.
  quantity: '*'
groups:
- name: meanings_tables
  doc: Group containing MeaningsTable objects that provide meanings for values
    in VectorData columns within this DynamicTable.
  quantity: '?'
  groups:
  - data_type_inc: MeaningsTable
    doc: MeaningsTable objects that provide meanings for values in VectorData
      columns within this DynamicTable. Tables should be named according to the
      column they provide meanings for with a "_meanings" suffix. e.g., if a
      VectorData column is named "stimulus_type", the corresponding
      MeaningsTable should be named "stimulus_type_meanings".
    quantity: '*'

3.3.6. AlignedDynamicTable

Extends: DynamicTable

Description: see Section 2.2.6

YAML Specification:

data_type_def: AlignedDynamicTable
data_type_inc: DynamicTable
doc: DynamicTable container that supports storing a collection of sub-tables.
  Each sub-table is a DynamicTable itself that is aligned with the main table by
  row index. I.e., all DynamicTables stored in this group MUST have the same
  number of rows. This type effectively defines a 2-level table in which the
  main data is stored in the main table implemented by this type and additional
  columns of the table are grouped into categories, with each category being
  represented by a separate DynamicTable stored within the group.
attributes:
- name: categories
  dtype: text
  dims:
  - num_categories
  shape:
  - null
  doc: The names of the categories in this AlignedDynamicTable. Each category is
    represented by one DynamicTable stored in the parent group. This attribute
    should be used to specify an order of categories and the category names must
    match the names of the corresponding DynamicTable in the group.
groups:
- data_type_inc: DynamicTable
  doc: A DynamicTable representing a particular category for columns in the
    AlignedDynamicTable parent container. The table MUST be aligned with (i.e.,
    have the same number of rows) as all other DynamicTables stored in the
    AlignedDynamicTable parent container. The name of the category is given by
    the name of the DynamicTable and its description by the description
    attribute of the DynamicTable.
  quantity: '*'

3.3.7. MeaningsTable

Extends: DynamicTable

Description: see Section 2.2.7

YAML Specification:

data_type_def: MeaningsTable
data_type_inc: DynamicTable
doc: A table to store information about the meanings of values in a linked
  VectorData object. All possible values of the linked VectorData object should
  be present in the 'value' column of this table, even if the value is not
  observed in the data. Additional columns may be added to store additional
  metadata about each value. The name of the MeaningsTable should correspond to
  the name of the linked VectorData object with a "_meanings" suffix. e.g., if
  the linked VectorData object is named "stimulus_type", the corresponding
  MeaningsTable should be named "stimulus_type_meanings".
datasets:
- name: value
  data_type_inc: VectorData
  doc: The value of a row in the linked VectorData object.
- name: meaning
  data_type_inc: VectorData
  dtype: text
  doc: The meaning of the value in the linked VectorData object.
links:
- name: target
  target_type: VectorData
  doc: Link to the VectorData object for which this table provides meanings.

3.4. Sparse data types

data types for different types of sparse matrices

3.4.1. CSRMatrix

Extends: Container

Description: see Section 2.3.1

YAML Specification:

data_type_def: CSRMatrix
data_type_inc: Container
doc: A compressed sparse row matrix. Data are stored in the standard CSR format,
  where column indices for row i are stored in indices[indptr[i]:indptr[i+1]]
  and their corresponding values are stored in data[indptr[i]:indptr[i+1]].
attributes:
- name: shape
  dtype: uint
  dims:
  - number of rows, number of columns
  shape:
  - 2
  doc: The shape (number of rows, number of columns) of this sparse matrix.
datasets:
- name: indices
  dtype: uint
  dims:
  - number of non-zero values
  shape:
  - null
  doc: The column indices.
- name: indptr
  dtype: uint
  dims:
  - number of rows in the matrix + 1
  shape:
  - null
  doc: The row index pointer.
- name: data
  dims:
  - number of non-zero values
  shape:
  - null
  doc: The non-zero values in the matrix.

3.5. Resource reference data types

data types for storing references to web accessible resources

3.5.1. HERD

Extends: Container

Description: see Section 2.4.1

YAML Specification:

data_type_def: HERD
data_type_inc: Container
doc: HDMF External Resources Data Structure. A set of six tables for tracking
  external resource references in a file or across multiple files.
datasets:
- name: keys
  data_type_inc: Data
  dtype:
  - name: key
    dtype: text
    doc: The user term that maps to one or more resources in the `resources`
      table, e.g., "human".
  dims:
  - num_rows
  shape:
  - null
  doc: A table for storing user terms that are used to refer to external
    resources.
- name: files
  data_type_inc: Data
  dtype:
  - name: file_object_id
    dtype: text
    doc: The object id (UUID) of a file that contains objects that refers to
      external resources.
  dims:
  - num_rows
  shape:
  - null
  doc: A table for storing object ids of files used in external resources.
- name: entities
  data_type_inc: Data
  dtype:
  - name: entity_id
    dtype: text
    doc: The compact uniform resource identifier (CURIE) of the entity, in the
      form [prefix]:[unique local identifier], e.g., 'NCBI_TAXON:9606'.
  - name: entity_uri
    dtype: text
    doc: The URI for the entity this reference applies to. This can be an empty
      string. e.g.,
      https://www.ncbi.nlm.nih.gov/Taxonomy/Browser/wwwtax.cgi?mode=info&id=9606
  dims:
  - num_rows
  shape:
  - null
  doc: A table for mapping user terms (i.e., keys) to resource entities.
- name: objects
  data_type_inc: Data
  dtype:
  - name: files_idx
    dtype: uint
    doc: The row index to the file in the `files` table containing the object.
  - name: object_id
    dtype: text
    doc: The object id (UUID) of the object.
  - name: object_type
    dtype: text
    doc: The data type of the object.
  - name: relative_path
    dtype: text
    doc: The relative path from the data object with the `object_id` to the
      dataset or attribute with the value(s) that is associated with an external
      resource. This can be an empty string if the object is a dataset that
      contains the value(s) that is associated with an external resource.
  - name: field
    dtype: text
    doc: The field within the compound data type using an external resource.
      This is used only if the dataset or attribute is a compound data type;
      otherwise this should be an empty string.
  dims:
  - num_rows
  shape:
  - null
  doc: A table for identifying which objects in a file contain references to
    external resources.
- name: object_keys
  data_type_inc: Data
  dtype:
  - name: objects_idx
    dtype: uint
    doc: The row index to the object in the `objects` table that holds the key
  - name: keys_idx
    dtype: uint
    doc: The row index to the key in the `keys` table.
  dims:
  - num_rows
  shape:
  - null
  doc: A table for identifying which objects use which keys.
- name: entity_keys
  data_type_inc: Data
  dtype:
  - name: entities_idx
    dtype: uint
    doc: The row index to the entity in the `entities` table.
  - name: keys_idx
    dtype: uint
    doc: The row index to the key in the `keys` table.
  dims:
  - num_rows
  shape:
  - null
  doc: A table for identifying which keys use which entity.