WCML

WCML is a library for outputting CML data. It wraps all the necessary XML calls, such that you should never need to touch any WXML calls when outputting CML.

The CML output is conformant to version 2.4 of the CML schema. The output can also be made conformant to the CompChem convention.

The available functions and their intended use are listed below. Quite deliberately, no reference is made to the actual CML output by each function.

Wcml is not intended to be a generalized Fortran CML output layer. rather it is intended to be a library which allows the output of a limited set of well-defined syntactical fragments.

Further information on these fragments, and on the style of CML generated here, is available at http://www.uszla.me.uk/specs/subset.html.

This section of the manual will detail the available CML output subroutines.

Use of WCML

wcml subroutines can be accessed from within a module or subroutine by inserting

 use FoX_wcml

at the start. This will import all of the subroutines described below, plus the derived type xmlf_t needed to manipulate a CML file.

No other entities will be imported; public/private Fortran namespaces are very carefully controlled within the library.

Dictionaries.

The use of dictionaries with WCML is strongly encouraged. (For those not conversant with dictionaries, a fairly detailed explanation is available at http://www.xml-cml.org/information/dictionaries)

In brief, dictionaries are used in two ways.

Identification

Firstly, to identify and disambiguate output data. Every output function below takes an optional argument, dictRef="". It is intended that every piece of data output is tagged with a dictionary reference, which will look something like nameOfCode:nameOfThing.

So, for example, in SIESTA, all the energies are output with different dictRefs, looking like: siesta:KohnShamEnergy, or siesta:kineticEnergy, etc. By doing this, we can ensure that later on all these numbers can be usefully identified.

We hope that ultimately, dictionaries can be written for codes, which will explain what some of these names might mean. However, it is not in any way necessary that this be done - and using dictRef attributes will help merely by giving the ability to disambiguate otherwise indistinguishable quantities.

We strongly recommend this course of action - if you choose to do follow our recommendation, then you should add a suitable Namespace to your code. That is, immediately after cmlBeginFile and before cmlStartCml, you should add something like:

call cmlAddNamespace(xf=xf, 'nameOfCode', 'WebPageOfCode')

Again, for SIESTA, we add:

call cmlAddNamespace(xf, 'siesta, 'http://www.uam.es/siesta')

If you don't have a webpage for your code, don't worry; the address is only used as an identifier, so anything that looks like a URL, and which nobody else is using, will suffice.

Quantification

Secondly, we use dictionaries for units. This is compulsory (unlike dictRefs above). Any numerical quantity that is output through cmlAddProperty or cmlAddParameter is required to carry units. These are added with the units="" argument to the function. In addition, every other function below which will take numerical arguments also will take optional units, although default will be used if no units are supplied.

Further details are supplied in section Units below.

General naming conventions for functions.

Functions are named in the following way:

Conventions used below.

Note that where strings are passed in, they will be passed through entirely unchanged to the output file - no truncation of whitespace will occur.

Also note that wherever a real number can be passed in (including through anytype) then the formatting can be specified using the conventions described in StringFormatting

Where an array is passed in, it may be passed either as an assumed-shape array; that is, as an F90-style array with no necessity for specifying bounds; thusly:

integer :: array(50)
call cmlAddProperty(xf, 'coords', array)

or as an assumed-size array; that is, an F77-style array, in which case the length must be passed as an additional parameter:

integer :: array(*)
call cmlAddProperty(xf, 'coords', array, nitems=50)

Similarly, when a matrix is passed in, it may be passed in both fashions:

integer :: matrix(50, 50)
call cmlAddProperty(xf, 'coords', matrix)

or

integer :: array(3, *)
call cmlAddProperty(xf, 'coords', matrix, nrows=3, ncols=50)

All functions take as their first argument an XML file object, whose keyword is always xf. This file object is initialized by a cmlBeginFile function.

It is highly recommended that subroutines be called with keywords specified rather than relying on the implicit ordering of arguments. This is robust against changes in the library calling convention; and also stepsides a significant cause of errors when using subroutines with large numbers of arguments.

Units

Note below that the functions cmlAddParameter and cmlAddProperty both require that units be specified for any numerical quantities output.

If you are trying to output a quantity that is genuinely dimensionless, then you should specify units="units:dimensionless"; or if you are trying to output a countable quantity (eg number of CPUs) then you may specify units="units:countable".

For other properties, all units should be specified as namespaced quantities. If you are using a very few common units, it may be easiest to borrow definitions from the provided dictionaries;

(These links do not resolve yet.)

cmlUnits: http://www.xml-cml.org/units/units
siUnits: http://www.xml-cml.org/units/siUnits
atomicUnits: http://www.xml-cml.org/units/atomic

A default units dictionary, containing only the very basic units that wcml needs to know about, which has a namespace of: http://www.uszla.me.uk/FoX/units, and wcml assigns it automatically to the prefix units.

This is added automatically, so attempts to add it manually will fail.

The contents of all of these dictionaries, plus the wcml dictionary, may be viewed at: http://www.uszla.me.uk/unitsviz/units.cgi.

Otherwise, you should feel at liberty to construct your own namespace; declare it using cmlAddNamespace, and markup all your units as:

 units="myNamespace:myunit"

Functions for manipulating the CML file:

This takes care of all calls to open a CML output file.

This takes care of all calls to close an open CML output file, once you have finished with it. It is compulsory to call this - if your program finished without calling this, then your CML file will be invalid.

This adds a namespace to a CML file.
NB This may only ever be called immediately after a cmlBeginFile call, before any output has been performed. Attempts to do otherwise will result in a runtime error.

This will be needed if you are adding dictionary references to your output. Thus for siesta, we do:

call cmlAddNamespace(xf, 'siesta', 'http://www.uam.es/siesta')

and then output all our properties and parameters with dictRef="siesta:something".

This pair of functions begin and end the CML output to an existing CML file. It takes care of namespaces.

Note that unless specified otherwise, there will be a convention attribute added to the cml tag specifying FoX_wcml-2.0 as the convention. (see http://www.uszla.me.uk/FoX for details)

Start/End sections

This pair of functions open & close a metadataList, which is a wrapper for metadata items.

This pair of functions open & close a parameterList, which is a wrapper for input parameters.

This pair of functions open & close a propertyList, which is a wrapper for output properties.

Start/end a list of k-points (added using cmlAddKpoint below)

Note that in most cases where you might want to use a serial number, you should probably be using the cmlStartStep subroutine below.

This pair of functions open & close a module of a computation which is unordered, or loosely-ordered. For example, METADISE uses one module for each surface examined.

This pair of functions open and close a module of a computation which is strongly ordered. For example, DLPOLY uses steps for each step of the simulation.

Adding items.

This adds a single item of metadata. Metadata vocabulary is completely uncontrolled within WCML. This means that metadata values may only be strings of characters. If you need your values to contain numbers, then you need to define the representation yourself, and construct your own strings.

This function adds a tag representing an input parameter

This function adds a tag representing an output property

Adding geometry information

Outputs an atomic configuration. Bonds may be added using the optional arguments bondAtom1Refs, bondAtom2Refs and bondOrders. All these arrays must be the same lenght and all must be present if bonds are to be added. Optionally, bondIds can be used to add Ids to the bond elements. Some valididity constraints are imposed (atomsRefs in the bonds must be defined, bonds cannot be added twice). The meaning of the terms "molecule", "bond" and "bond order" is left loosly defined.

Outputs information about a unit cell, in lattice-vector form

Outputs information about a unit cell, in crystallographic form

Adding eigen-information

Start a kpoint section.

End a kpoint section.

Add an empty kpoint section.

Start a section describing one band.

End a section describing one band.

Add a single eigenvalue to a band.

Add a list of eigenvalues for a kpoint

Add a phononic eigenpoint to the band - which has a single energy, and a 3xN matrix representing the eigenvector.

Echoing input files

It is often considered useful to include a direct representation of input data within an applications output files. FoX_wcml contains a number of procedures to allow this in a CML document based on a specification described in a manuscript currently in review (de Jong, Walker and Hanwell "From Data to Analysis: Linking NWChem and Avogadro with the Syntax and Semantics of Chemical Markup Language". The approach is also designed to make data recovery using an XSL transform straightforward. File metadata such as the original filename is also accessible. We assume that only ASCII data must be stored, arbitrary binary files are out of scope as are XML documents and non-ASCII textural data. Two methods are provided with the most appropriate being dependent on the design of the application.

Using file names

In the this approach the single subroutine, cmlDumpInputDec, is called with an array of file names as input arguments. In turn each file is opened, its contents are written to the CML document in the appropriate form, before the file is closed.

Line by line

Start an outer wrapper for input data.

End the outer wrapper.

Start a wrapper for a single "file" or similar concept.

End the file wrapper.

Put a line of text into the file wrapper.

The convoluted nature of file handling in Fortran combined with the way that some applications read their input data means that this approach is not always available (for example, if the input file is held open for the duration of the calculation, or if data is read from standard input) so an alternative interface with five subroutines (cmlStartDecList, cmlStartDec, cmlAddDecLine, cmlEndDec and cmlEndDecList) is provided. These must be called in order (and usually in two loops, one over files, and an inner loop over lines in each file).

Common arguments

All cmlAdd and cmlStart routines take the following set of optional arguments: