1 Introduction
2 ------------
3
4 XSLTools is a collection of modules and packages facilitating the development
5 of applications based on XML, XSL stylesheets and transformations, notably Web
6 applications involving complicated Web forms potentially consisting of
7 editable hierarchical structures and potentially involving "live" or "in-page"
8 dynamic updates to portions of those Web forms.
9
10 Quick Start
11 -----------
12
13 Try running the demo:
14
15 python tools/demo.py
16
17 An introductory guide to creating applications can be found in the docs
18 directory - see docs/index.html for the start page.
19
20 Contact, Copyright and Licence Information
21 ------------------------------------------
22
23 The current Web page for XSLTools at the time of release is:
24
25 http://www.boddie.org.uk/python/XSLTools.html
26
27 Copyright and licence information can be found in the docs directory - see
28 docs/COPYING.txt, docs/LICENCE.txt and docs/LICENCE-Sarissa.txt for more
29 information.
30
31 Dependencies
32 ------------
33
34 XSLTools has the following basic dependencies:
35
36 Package Release Information
37 ------- -------------------
38
39 libxml2dom 0.2.1
40 libxml2 Tested with 2.6.17
41 libxslt Tested with 1.1.12
42
43 The example Web applications require WebStack (release 1.0 or later).
44 The example PyQt applications have been tested with PyQt 3.15.
45
46 New in XSLTools 0.2 (Changes since XSLTools 0.1)
47 ------------------------------------------------
48
49 * Made a new XSLTools package and moved XSLOutput into it.
50 * Improved serialisation of transformation results so that output options
51 are observed (in some cases, at least).
52 * Fixed stylesheet and reference document paths so that libxslt should not
53 now become confused by ambiguous relative paths.
54 * Added expression parameters to XSLOutput.Processor so that in-document
55 data can be used to, for example, initialise multiple-choice field values.
56 * Added input/initialiser support so that input documents can be tidied or
57 initialised using information from the template.
58 * Added template:init for use with template:element in XSLForms to control
59 element initialisation where necessary.
60 * Added special high-level "macro" attributes (eg. template:attribute-field)
61 which should make templates easier to write and maintain.
62 * Added template:if to XSLForms, providing conditional output of annotated
63 elements.
64 * Added set_document to XSLForms.Fields.Form.
65 * Added prepare_parameters to XSLForms.Resources.XSLFormsResource.
66 * Added element-path and url-encode XSLForms extension functions.
67 * Improved Unicode support in the XSLForms extension functions.
68 * Changed in-page requests to contain proper POST data.
69 * Updated the code to work with WebStack 1.0 changes and adopted the
70 new-style WebStack demonstration mechanism.
71 * Added XMLCalendar and XMLTable (to the XSLTools package).
72 * Added a dictionary (or word lookup) example application.
73 * Added a job candidate profile (or CV editor) example application.
74 * Added a template attribute reference and an XSLFormsResource guide to the
75 documentation.
76 * Added Debian package support.
77 * Added missing COPYING.txt file.
78 * Renamed the scripts to avoid naming issues in system-wide installations.
79 * Added a PyQt example based on the system configurator example, with the
80 form prepared in Qt Designer.
81
82 Notes on In-Page Update Functionality
83 -------------------------------------
84
85 Special note #1: Konqueror seems in certain cases to remember replaced form
86 content (when replaceChild is used to replace regions of the page which
87 include form elements). This causes the browser to believe that more form
88 fields exist on the page than actually do so, and subsequent form submissions
89 thus include the values of such removed fields. A special hack is in place to
90 disable form fields by changing their names, thus causing Konqueror to not
91 associate such fields with the real, active fields; this hack does not seem to
92 cause problems for Mozilla. This needs some investigation to determine in
93 exactly which circumstances the problem arises.
94
95 Special note #2: Konqueror also seems to crash if asked to find elements using
96 an empty 'id' attribute string. This needs some investigation to see if it
97 really is the getElementById call that causes the crash.
98
99 Special note #3: Konqueror's XMLHttpRequest seems to append null characters to
100 the end of field values. Attempting to prune them before the request is sent
101 fails with a function like the following:
102
103 function fixValue(fieldValue) {
104 if (fieldValue.length == 0) {
105 return fieldValue;
106 } else if (fieldValue[fieldValue.length - 1] == '\0') {
107 return fieldValue.substr(0, fieldValue.length - 1);
108 } else {
109 return fieldValue;
110 }
111 }
112
113 This may be because it is the entire message that is terminated with the null
114 character, and that this happens only upon sending the message. Consequently,
115 some frameworks (notably mod_python) do not support in-page functionality when
116 used from Konqueror.
117
118 Various browsers (eg. Mozilla/Firefox, Konqueror) will not allow the
119 XMLHttpRequest in-page updates to function unless the URL used in the
120 requestUpdate JavaScript function is compatible with the URL at which the
121 browser finds the application. Currently, relative URLs are in use to avoid
122 this issue of compatibility, but should an absolute URL be deduced using the
123 WebStack API and then used, it may be possible that the values returned by
124 that API do not match the actual addresses entered into the address bar of the
125 browser.
126
127 To check the behaviour of the applications, it is possible to view the
128 document source of the pages served by applications and to verify that the
129 URLs mentioned in the JavaScript function calls (to 'requestUpdate') either be
130 a relative link or involve a URL similar to that which appears in the
131 browser's address bar. In some environments, the use of 'localhost' addresses
132 often confuses the browser and server; one workaround is to use real host
133 names or addresses instead of 'localhost'.
134
135 Choosing an element-path:
136
137 When specifying the "context" of the in-page update, one must imagine which
138 element the template fragment should operate within. If the template:id
139 attribute marks a particular section, then the element-path should be a path
140 to the applicable context element for that section in the complete template
141 document. Note that if a template:element attribute appears on the same
142 element as the template:id attribute then the element-path should refer to the
143 element specified in the template:element attribute.
144
145 Choosing where to put template:attribute, template:id and id:
146
147 When specifying the extent of a template fragment, one must be sure not to put
148 the template:id attribute on the same element as a template:attribute
149 annotation; otherwise, the generated code will be improperly extracted as a
150 fragment producing two versions of the element - one for when the specified
151 attribute is present, and one for when it is not present. Generally,
152 template:id and id can be placed on the same node, however.
153
154 Stable element ordering and element-path:
155
156 Within the element-path, the numbering of the elements will start at 1.
157 Therefore it is vital to choose a region of the form data structure with the
158 element-path which is isolated from surrounding elements whose positions would
159 otherwise be dependent on a stable ordering of elements, and whose processing
160 would be disrupted if some new elements suddenly appeared claiming the same
161 positions in the document. For example:
162
163 <item value=""> .../item$1/value
164 <type value=""/> .../item$1/type$1/value
165 <comment value=""/> .../item$1/comment$2/value
166 </item>
167
168 In-page update...
169
170 <comment value=""/> .../item$1/comment$1/value
171
172 Notes on XSL
173 ------------
174
175 libxslt seems to be quite liberal on the definition of runtime parameters, in
176 that there is no apparent need to explicitly declare the corresponding global
177 variables in stylesheets. Whilst this is nice, we may eventually need to
178 detect such variables and add them in the preparation process.
179
180 Release Procedures
181 ------------------
182
183 Update the XSLTools/__init__.py and XSLForms/__init__.py __version__
184 attributes.
185 Change the version number and package filename/directory in the documentation.
186 Change code examples in the documentation if appropriate.
187 Update the release notes (see above).
188 Check the setup.py file and ensure that all package directories are mentioned.
189 Check the release information in the PKG-INFO file and in the package
190 changelog (and other files).
191 Tag, export.
192 Generate the API documentation.
193 Remove generated .pyc files: rm `find . -name "*.pyc"`
194 Archive, upload.
195 Upload the introductory documentation.
196 Update PyPI, PythonInfo Wiki, Vaults of Parnassus entries.
197
198 Generating the API Documentation
199 --------------------------------
200
201 In order to prepare the API documentation, it is necessary to generate some
202 Web pages from the Python source code. For this, the epydoc application must
203 be available on your system. Then, inside the distribution directory, run the
204 apidocs.sh tool script as follows:
205
206 ./tools/apidocs.sh
207
208 Some warnings may be generated by the script, but the result should be a new
209 apidocs directory within the distribution directory.
210
211 Making Packages
212 ---------------
213
214 To make Debian-based packages:
215
216 1. Create new package directories under packages if necessary.
217 2. Make a symbolic link in the distribution's root directory to keep the
218 Debian tools happy:
219
220 ln -s packages/ubuntu-hoary/python2.4-xsltools/debian/
221
222 3. Run the package builder:
223
224 dpkg-buildpackage -rfakeroot
225
226 4. Locate and tidy up the packages in the parent directory of the
227 distribution's root directory.
