About the Dataset document

In DAP2 and DAP4, a dataset is defined as a collection of variables, each of which is completely described by a tuple that consists of a name, type and value(s) along with a hierarchical collection of 'attributes' that themselves are made up of name-type-value(s) tuples. The variables are organized in a hierarchy.

The Dataset document is a text/xml representatin of those variables and their organiation within the dataset.

A note about XML element names:

Element names that start with capital letters correspond to parts of the DAP4 data model while those that start with lowercase letters are used for document structure and syntax.

About changes in going from DAP2 to DAP4

DAP 4.0 introduces SharedDimension, Group, Opaque, 64-bit integers and UnsignedByte. In addition: The syntax for Array has been changed so that it's easier for processing software to figure out the type of an array; Grids have been generalized so that there can be any number of 'Array' parts (and the Maps may be multi-dimensional); and the Attribute type OtherXML has been made its own element (it's no longer a type of 'Attribute').

These elements are used in several places to hold the 'semantic' and/or 'use' metadata for the dataset, its groups and variables. ( OtherXML*, Attribute* )

Dataset

This is the XML representation of a data source in DAP 4.

Note that the 'blob' element is only present when this is used as the prefix for the Data response.

Grammar: ( OtherXML*, Attribute*, Group+, blob? )

Element attributes:

name
The name of the data source; often a string used to uniquely reference the data source wrt a particular server
dapVersion
The protocol version that corresponds to this document.
xml:base
The URL the references the DAP4 service endoint used to access this dataset.
 DAP uses attributes as a way to encode information that data providers have bundled with data sources. This element is recursive. Each Attribute element defines a lexical scope. If there are no nodes present the type must be "Container". name: The name of the attribute; must be unique within the scope. type: The type of the attribute. Attributes are limited to simple types, vectors of simple types and 'Containers' which are essentially structure types. namespace: Use this to indicate that the given attribute means the same thing as the matching item in the given namespace. This optional attribute is here to help preserve information that a data server might know to be true and that a client application could not assume with certainty. ( value* | ( OtherXML*, Attribute* ) ) Changes from DAP2 to DAP4: The Byte type is now signed and unsigned bytes are now represented by the 'UByte' data type. The types Int64 and UInt64 represent signed and unsigned 64-bit integers. String uses UTF-8 in DAP4. Enumerations are now in the mix. Use this to embed arbitrary XML in a DDX. This functions like an Attribute and appears in the same places as an Attribute, but its contents are ignored by DAP software. Other software might find the information useful. The XML elements must satisfy the requirements for 'lax' processing under schema 1.0. In practice, that means just about anything. Using ##other versus ##any means that the enclosed XML MUST declare its namespace(s) and each element must be in a ns other than dap4's. Of course, the namespace(s) used by the XML might be declared elsewhere in the doc. name: A name to associate with this chunk of XML *: This element can contain any other attributes that conform to the schema 1.0 definition of 'lax' processing ( xs:any+ ) A Group is a lexical scoping tool used to replicate HDF5 and netCDF4 Groups. Each Group defines a lexical scope. Each dataset has at least one Group; if only one is present, it may be anonymous. In this case, by convention, it's name attribute should be 'anonymous', the default value. name: The name of the Group ( OtherXML*, Attribute*, Dimension*, ScalarType+ ) This defines a dimension (a name and size) that may be shared between Grids and/or Arrays. name: The name of the dimension size: The size of the dimension This defines the values of an enumeration. name: The name of the dimension type: The size of the dimension DAP cardinal data types This provides a version of BaseType that does not allow child elements since Grid and Sequence may not be array types. This extended ScalarBaseType so that instances may be either scalar or N-dimensional. This extended BaseType so that instances must be N-dimensional (N >= 1) and reference one or more SharedDimension objects. Grids cannot have scalars (they have only Arrays and Maps) and Arrays have both dimensions and references to maps. Regular arrays do not have references to maps. This attribute is used to reference a SharedDimension object. name: The name of the dimension size: The size of the dimension ref: A reference to a Dimension definition Note: either name and size must be present or only ref must be present name: Name for this Map. Maps are associated with Array dimentions by name, so this is a required attribute. type: The type of a Map is limited to the Cardinal types. NB: Note the limitation on the type of a Map, which excludes Maps that are Opaques, Structures, Sequences or Grids. ( OtherXML*, Attribute*, dimensions ) A Grid is a type that relates one or more Maps (aka coordinate variables) to the dimensions of one or more Array variables. It is often the case that Maps correspond to independent variables like Latitude or sample number and Arrays represent dependent variables. Note that Map elements either specify a name and size or reference a Dimension. The scope in which the Dimension can be located is limited to an enclosing Group (but it is not limited to the immediate parent Group). name: The name of the variable ( OtherXML*, Attribute*, Array+, Map+) A Structure; a simple aggregation of variables. Unlike a Group, it's possible to from an Array of Structures. name: The name of the variable ( OtherXML*, Attribute*, ( Byte | ... | Grid )+ ) A Sequence is a type that holds tabular data where each row of the table represents a relation, as in a relational database. Sequences can nest, but Arrays of Sequences are not supported. name: The name of the variable ( OtherXML*, Attribute*, ( Byte | ... | Grid )+ )

The 'Blob' element is used to point to an associated data document. When DAP is used to access metadata only for a data source, no 'blob' element will be present. However, when a request for data is made, the Dataset element holds a description of the data values and the blob points to a place where those values will be found. In DAP4 the Blob element refers to binary (application/octet-stream) part within a multi-part MIME document or it refers to a separate document, possibly not imediately available. The latter case is included to support asynchronous responses (i.e., responses that cannot be returned quickly). See http://www.w3.org/TR/xlink11/.

The dc:date child element

If present, the element indicates the time or time range when the information will be available. This is only sensible (i.e., valid) when the response is asynchronous (i.e., xlink:role="http://xml.opendap.org/dap/DAP4#asynchronous"). Dublin core defines both a date and a date range. In the case of a date, this element indicates when the information will likely be available; when its value is a range, it denotes the when it will likely be present and when it wil go away.

Examples of the element:

  • <dc:date>1994-11-05T13:15:30Z</dc:date>
  • <dc:date>1994-11-05T13:15:30Z/1994-11-06T00:00:00Z</dc:date>

Element attributes:

xlink:href:
Refers to the multi-part MIME document part that holds the data values, encoded using XDR or to a separate document. In the first case, the IRI must begin with 'cid:' (see the owsManifest schema for an example of this use; http://schemas.opengis.net/ows/2.0/owsManifest.xsd). In the latter case, the IRI must refer to a remote resourse, and will likely start 'http:'.
xlink:type:
Always 'simple'.
xlink:role:
If present the only values DAP4 supports are 'asynchronousResponse' and 'synchronousResponse'.