Survey schema tools for Socrates.
The schema is defined in survey.xsd. To test your survey model against the schema, you might use xmllint like so:
xmllint --schema survey.xsd --noout <your xml model>
The survey consists of four (main) parts:
- data
- model
- layout
- submission
For i18n the new survey model uses the schema for i18n by Zope:
http://xml.zope.org/namespaces/i18n
This allows the i18n:trans attribute on translatable items like label, hint, alert, etc.
The data part defines the variables (or: fields) that will hold the collected data. So, for instance, if you'd need to know someone's address, the data could be:
<data>
<var name="address" />
<var name="city" />
<var name="zipcode" />
</data>
Variables can be grouped together, to create a hierarchy or grouping:
<data>
<vargroup name="address">
<var name="street" />
<var name="city" />
<var name="zipcode" />
</vargroup>
</data>
Groups can contain other groups, so nesting is unlimited. This can be used to address groups of variables in one go. See the section on the model for more details.
A variable can also hold a default value:
<var name="zipcode">9999..</a>
"It's only a model!" "Shhhh!"
The model defines properties of the data, like whether a variable must have a value for the data to be valid, or specifying the type of the variable. The following properties may be set (in the order of the list below):
- relevant
- required
- readonly
- calculate
- constraint
- datatype
- default
N.B order is important!
All properties except for the datatype can be specified in terms of an expression, like:
relevant="true()"
required="foo == 1"
To apply properties to variables, 'binding' is used. This is done through expressions that specify to what variables the properties apply, using regular expressions:
<properties id="required">
<bind>/address/city</bind>
<required>true</required>
</properties>
or:
<properties id="required">
<bind>/address/.*</bind>
<required>true</required>
</properties>
You can specify as many bind tags on a properties element as you like.
Specify whether the variable is relevant for the complete data. For example whenever someone filling in a survey says he has had no education, a further question as to what type of school he/she went to is irrelevant. Defaults to 'true'.
Specify whether some variable is required for the data to be valid. Defaults to 'false'.
Specify whether some variable cannot be set. Defaults to 'false'.
Constrain possible values of a variable, like 'smaller than' or 'in the range of ...'.
Calculate the value of a variable.
Specify the datatype of the variable. This places extra constraints on the variable, but may also affect display.
Specify a default value, using an expression. The default is calculated at the moment the variable is accessed for the first time.
The view part of the survey specifies what it 'looks like', although in theory the survey could also be converted to automated speech...
The layout specifies what kind of control is used to obtain data, so for instance a select form element to present a list of values to the user for the question "What is your favourite color?". A control is bound to the variable using the same bind attribute as for properties. The semantics of
<select id="favcolor" bind="fav-colr">
...
</select>
is therefore: store the answer in the fav-colr variable.
The layout section enables grouping of control elements into pages, or fieldsets, etc. The grouping also enables you to say something about the layout of the group, using the layout attribute.
So a group:
<group id="some group" layout="page">
...
</group>
will render all it's listed controls on a page, wheres a group like:
<group id="some group" layout="flow">
<property name="dir">h</property>
...
</group>
will render it's controls in a horizontal flow.
Specify the way the survey data should be handled.
The survey model allows for internationalisation (i18n), or in other words, it supports translation information. All elements that are translatable should hold a i18n:translate attribute. Elements that are translatable typically are question labels, hints, option labels, etc.
The translation definition assumes an actual translation mechanism like gettext, which is based on message id's. The message id is the key for the translation into different languages.
If the value of the i18n:translate attribute is empty, the content of the element is taken as the message id:
<label i18n:translate="">What is your favourite color?</label>
=> message id = 'What is your favourite color?'
If the i18n:translate attribute actually holds a value by itself, this is taken as the message id:
<label i18n:translate="fav_color">What is your favourite color?</label>
=> message id = 'fav_color'