DAP 4.0 Design

This is a Draft

Introduction

Overall Operation

Definitions

Type Definition

The representation of something in a data set which defines a data type; types are defined in a data set. The DAP provides a way to represent this definition and use it as a stand-in for a definition built using the DAP-supplied types.

Dimension

A name bound to a size, e.g., "lon" has a size of 1024

Coordinate Variable

A name bound to both a dimension and a data type, e.g.,"height" is a vector of dimension "height_size" 32-bit floating point numbers, or "latitude" is an array of dimension "x" by dimension "y" 32-bit floating point numbers.

Grid

One or more N-dimensional arrays of values bound to 1 to N coordinate variables.

Data Model

DAP is evolving to meet the increasing complexity of data sources and the improving capabilities of analysis software to act as clients for various online data sources. Part of this evolution is to adopt the Common Data Model that has been developed by Unidata and other changes are the introduction of new data types and the expansion of capabilities of the widely used 'Grid' type. Finally, some types like Array of Grid and Array of Sequence have been eliminated because they are hard to implement. By eliminating them we hope client authors will focus on complete implementations of the existing types.

The DAP 2 and DAP 4 data models are shown to the right.

High resolution version of the DAP 2 Data Model High resolution version of the DAP 4 Data Model

Data Types

DAP 4 will have a small increase in supported data types. All of the DAP 2 data types describes in ESE RFC 004.11 will be supported with their existing definition with the exceptions that Grid will be expanded so that it can be used in more situations and strings will comply with UTF-8. The additional types will support 64-bit integers, an Opaque type that can be used for data objects like JPEG images, Groups that can be used to build logical collections as in NetCDF4 or HDF5 (with some limitations over HDF5's definition of Group). In addition, the server-side of DAP 4 will provide for Shared Dimensions and Type Definitions. Both Shared Dimensions and Type Definitions will allow data systems that have these to be presented with better fidelity than DAP 2.

New Datatypes

Type Definitions

Type Definitions provide a way for a data source to define a type using the any legal combination of the DAP types built-in tyeps and then define variables using that new name. Note: Since this facility does not provide a mechanism to define new operations it is technically a 'type equivalence' and not a 'type definition.'

New type definitions may appear only at the data source scope.

New type definitions may have attributes. Those attributes are bound to the type definition itself and are not inherited by variables that use the type definition.

Type definitions can be used for attributes. This opens up the whole question of allowing attributes to be a variable, et c., and all that it implies

The syntax for a type definition

<typedef name="point">
    << attributes >>
    <structure name="point">
        <int32 name="x"/>
        <int32 name="y"/>
    </structure>
</typedef>

Notes:

1. The contents of the <typedef> element are a DDX type
2. The enclosed type and the name become synonymous.
3. The type definition itself can include (optional) attributes.
4. The new type name can appear anywhere one of the built-in types can appear.

Use of a type definition, assuming the typedef above

<variable name="points" type="point">
    <dim size="100>
</variable>

or

<variable name="points">
    <type name="point"/>
    <dim size="100"/>
</variable>

Notes:

1. The XML elements <variable ...> and/or <type ..> are used because the typedef name can be any legal XML attribute value (a requirement because names in HDF5 are essentially not restricted to a particular character set) and thus cannot be an XML element name.
2. This example uses the new (proposed) syntax for array declarations.
3. Right now instead of a <variable ...> element in the DDX we have a collection of things like <Grid...>, <Array...> but we cannot reliably introduce arbitrary element names because of the character restrictions XML places on them. So I'm suggesting that we change the DDX to use the <variable...> element everywhere. And adopt the notion that A variable is scalar or vector/array depending on whether the element includes one or more <dim ...> elements.

Groups

The DDX will be modified so that it contains one or more Groups. If only one Group is present (which describes the case for DAP 3.2 and earlier) then the declaration can be left out, but if there are two or more groups, the declarations must be present.

Group characteristics:

• Any configuration of Groups other than one (anonymous) Group which holds all the variables in a data set must be declared.
• If declared, Groups must be named.
• A Group can contain any object, including a Group
• Variables and Attributes are named using / <group name> / ... / <variable name> to reflect their hierarchy.
• Each Group declares a new lexical scope for values.
• A Group cannot be an Array or a Grid (although the distinction between those two might become blurred or non-existent; Group is fundamentally a scalar container-type).
• This definition does not completely subsume the HDF5 Group type but is equivalent to the netCDF 4 version of it.

Shared Dimensions

Shared dimensions will be added to DAP in the dimensions section of the Dataset or Group objects. Each dimension will consist of a name and a size.

 
<dimension name="lat" size="1024"/>
<dimension name="lon" size="1024"/>

Characteristics of dimensions:

• Dimensions are not associated with a data type.
• Dimensions do not have attributes.
• Dimensions bound to a type define coordinate variables.
• Shared dimensions may be used by both Grids and Arrays.
• A Coordinate variable is not restricted to to being a single dimension; a N-dimensional coordinate variable would use N dimensions.

Background on shared dimensions and coordinate variables

From an email exchange, John Caron wrote:

James:

Is it that an dimension is a formal declaration of an independent parameter?

John:

I know that some people prefer that interpretation. My own opinion is that's it more complicated.

Abstractly, I think its reasonable to say that the number of dimensions of a variable indicates its dimensionality in the topological sense. I think its necessary to allow "independent variables" to have topological dimensionality > 1. eg lat(x,y), lon(x,y). lat and lon can still be considered independent variables, but they are not orthogonal. Neither is associated exclusively with one

dimension.

Concretely, dimensions are used for all sorts of reasons, and are not just about topological dimensionality. For instance, they control the grouping of data and the layout of files. So in real files, you see this mixture of uses.

That's why the explicit assignment of coord variables is needed, which makes your Grid attractive, because that's a way of explicitly saying what the independent variables are. One needs shared dimensions between data and coordinate variables, so that one can unambiguously assign coordinate values to a data value.

The downsides of using Grid for this purpose:

• the name "Grid" connotes gridded data, eg model data, and this shared dimension thing is needed for other types of data, eg point data.
• If Grid scopes the dimension, then all variables sharing a dimension have to be contained in the grid. So its impossible to have some dimensions globally shared, and others locally shared.

So my preference would be to use Groups to scope shared dimensions, rather than Grids. But still use Grids (or some evolution of Grids) to assign coordinate variables to data variables.

Opaque

The Opaque type is use to hold objects like JPEG images and other BLOB objects with significant internal structure that might be understood by some clients (e.g., an image display program) but which would be very cumbersome to describe using DAP's built-in types. Defining a variable of type 'Opaque' does not communicate any information about its content, although an attribute could be used to do that.

1. A variable of type Opaque is treated as a Byte array for the purposes of transmission. This means there is no attempt to re-order four-byte words to or from network byte order and that the block of bytes is extended to fill a four-byte boundary
2. The size of an Opaque variable is unknown until the data are read/received
3. The Opaque type should be considered similar to a 'simple type' in that it can appear in a Structure.
4. It's possible to have Arrays of Opaque variables

Questions:

1. Does it make sense to have a Grid of Opaque items?
2. Opaques in a Sequence?
3. Because of the complexity of dealing with variable length items, should client be allowed (they'd still be consider fully compliant clients) to not read Opaque variables?

64-bit Integers

Signed and Unsigned 64-bit integers.

Support for Existing Types

Changes in the Definition of Grid

While dimensions are scoped at the Dataset or Group level, coordinate variables are defined at the level of a Grid object. Grid objects in DAP4 are different from those in DAP2 in three ways beyond using (shared) dimensions:

1. Each Grid object may hold more than one Array (what is often a dependent variable);
2. Maps (often independent variables) may have more than one dimension; and
3. Each Array within a Grid is not constrained to use all of the Grid's Maps (aka coordinate variables).

N.B: Coordinate variables in a Grid object are called Maps to conform to the old nomenclature and to avoid (re)using the word dimension.

Features of the DAP4 and DAP2 Grid object:

1. Each Grid object defines a lexical scope.
2. There is an explicit relation between the Grid object's maps (coordinate variables) and the indicial extents of the array.

A very simple Grid object

<dataset name="example_1">

    <dimension name="lon" size="1024"/>
    <dimension name="lat" size="1024"/>

    <grid>
        <map name="lon" dim="lon" type="Float32"/>
        <map name="lat" dim="lat" type="Float32"/>

        <array name="SST">
            <Byte/>
            <map name="lon">
            <map name="lat">
        </array>
    </grid>

</dataset>

Notes:

1. The map object may have the same name as a dimension object.
2. Map objects may have attributes, even though they are not shown in the example.
3. In an Grid's array object, <map...> elements are used to specify the array's dimensions; the word dimension is avoided to cut down on confusion.

A more complex Grid object

<dataset>
    <dimension name="pt" size="4096">

    <grid>
        <map name="longitude" dim="pt" type="Float32"/>
        <map name="latitude" dim="pt" type="Float32"/>
        <map name="altitude" dim="pt" type="Float32"/>
        <map name="time" dim="pt" type="Float32">
            << attributes >> <!-- The syntax for attributes is in flux -->
        </map>

        <array name="Radioactivity">
            << attributes >> <!-- for example, scale_factor and add_offset -->
            <Byte/>
            <map name="longitude"/>
            <map name="latitude"/>
            <map name="altitude"/>
            <map name="time"/>
        </array>

        <array name="surface_temp">
             << attributes >>
             <float64/>
             <map name="longitude"/>
             <map name="latitude"/>
             <map name="time"/>
        </array>
    </grid>
</dataset>

An example Grid with Maps that are not vectors

<dataset>
    <dimension name="x" size="4096">
    <dimension name="y" size="4096">

    <grid name="SST_Swath">
        <!-- We could list multiple dims in a space-separated list
             but purists will gag. I'm experimenting with different 
             syntaxes -->
        <map name="longitude" type="Float32"/>
            <dim name="x"/>
            <dim name="y"/>
        </map>
        <map name="latitude" type="Float32"/>
            <dim name="x"/>
            <dim name="y"/>
        </map>

        <!-- This grid has two maps, each of which are two-dimensional
             arrays. It can be used to store satellite 'swath' data. -->
        <array name="SST">
            << attributes >> <!-- for example, scale_factor and add_offset -->
            <Byte/>
            <map name="longitude"/>
            <map name="latitude"/>
        </array>
    </grid>
</dataset>

Note:

1. The highest dimension of the Grid's Maps cannot exceed the dimensionality of the Grid's Array.
2. When using the [] operator on a Grid in a DAP Constraint expression, the arguments enclosed in the square brackets correspond to the dimensions declared in the Map and not the Maps themselves. Thus a CE like SST_Swath[10:20][40:50] means that the array SST_Swath.SST and the maps SST_Swath.longitude and SST_Swath.latitude will all be returned sub-sampled to elements 10 to 20 in their first dimension and 40 to 50 in their second. In a DAP2 grid where all of the maps are vectors, there is a one-to-one correspondence between the [] operators and Maps, but in a DAP4 Grid there is a one-to-one correspondence between the [] operators and dimensions.

Suggested Types not Included

Discussed in this section are types that are present in some other systems (e.g., ASN 1.1) but that are not includes in DAP 4 along with the rationale of not including them.

Enumeration

This type will be taxing to client builders because there is an unbounded set of potential values. At the same time most (all?) clients will implement this type as a set with no actual knowledge of the set elements semantics, so it is no different than a byte or integer type with a attribute that provides a binding between (integral) formal values and String, et c., actual values.

Boolean

The additional semantic information provided by this type seems very limited given the cost associated with each additional data type in a protocol such as DAP.

Date/Time

Of all the types suggested but not included, this has the most potential. Unfortunately, it's very unlikely that this type would be implemented correctly by a majority of servers. It is certainly possible to include it in the DAP itself, but servers would have to provide a mapping from the encoding of date/time in each relevant data source to some sort of a standard representation (e.g., ISO 8601). That seems unlikely and thus it seems most likely that a date/time type would not be used consistently. A better solution is to use Attributes which provide the potential for third party mediation or augmentation.

Attributes

Attributes in DAP 4 are largely unchanged from DAP 2 with the only change being the addition of a new type of attribute to hold XML which is supplied for a data source using some external system.

Services

We need to add some 'discovery' services so that a client can find out about different handlers and the customizations they support. One form of service discovery would be to return the server-side functions present on the server in a way that would lead to a client being able to ask for more information about each one, with the intent that a person could use this or that an interface could be built around this information (choose the function from a list). A second service would be to learn about installed handlers so that a client could then ask for information about a specific one. For example, a client might ask about handlers, see that a HDF5 handler is present and then ask for documentation on it. That would reveal that the handler encodes type definitions in global attributes - information that would be useful if you're trying to write a HDF5 file and want high fidelity.

The problem with services like this is that they can be a real pain if server installers have to write it up. These services need to be returning information that is part of the handler or function. Then the server interagates the handler or function and builds up the information. So the server installer does not have to write this stuff up.

Responses

Persistent representations

DDX Document Organization

DAP and the DDX will be extended to include Groups, Shared dimensions and user-defined types. Groups will be added as a kind of constructor-type with properties similar to Structure and to Java or C++ namespaces. Unlike Structure, Groups cannot be dimensioned.

A rough syntax which describes how these additions will fit into the DAP and the existing DDX Notation is (Replace with XML schema):

Dataset :== Groups
Groups :== null | Group Groups
Group :== Types Dimensions Attributes Variables Groups
Types :== null | Type Types
Dimensions :== null | Dimension Dimensions
Attributes :== null | Attribute Attributes
Variables :== null | Variable Variables

This pseudo-grammar does not capture what can be produced for a Group, et cetera. Instead it shows how these sections of the DDX must be organized. It also does not show that a valid Dataset can have only Types (user-define types) and does not need to have variables, but it must have one or the other or both.

Examples

Group Examples

This data set contains one Group - the root group - which has by convention the name '/'

<Dataset ... >
    ...
</Dataset>

This data set contains two Groups, one after the other.

<Dataset ... >
<group name="primary">
    ...
</group>
<group name="secondary">
    ...
</group>
</Dataset>

This data set contains more Groups, and shows they can be nested.

<Dataset ... >
<group name="primary">
    ...
    <group name="in_situ">
        ...
    </group>

</group>
<group name="secondary">
    ...
</group>
</Dataset>