poi/src/documentation/xdocs/hssf/record-generator.xml

115 lines
4.7 KiB
XML

<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE document PUBLIC "-//APACHE//DTD Documentation V1.0//EN" "../dtd/document-v10.dtd">
<document>
<header>
<title>Record Generator HOWTO</title>
<authors>
<person email="glens@apache.org" name="Glen Stampoultzis" id="glens"/>
<person email="acoliver@apache.org" name="Andrew C. Oliver" id="acoliver"/>
</authors>
</header>
<body>
<s1 title="How to Use the Record Generator">
<s2 title="History">
<p>
The record generator was born from frustration with translating
the Excel records to Java classes. Doing this manually is a time
consuming process. It's also very easy to make mistakes.
</p>
<p>
A utility was needed to take the defintition of what a
record looked like and do all the boring stuff. Thus the
record generator was born.
</p>
</s2>
<s2 title="Capabilities">
<p>
The record generator takes XML as input and produced the following
output:
<ul>
<li>A Java file capabile of decoding and encoding the record.</li>
<li>A test class with provides a fill-in-the-blanks implementation of a test case
for ensuring the record operates as designed.</li>
</ul>
</p>
</s2>
<s2 title="Usage">
<p>
The record generator is invoked as an Ant target (generate-records). It goes
through looking for all files in src/records/defintitions ending with _record.xml.
It then creates two files; the Java record definition and the Java test case template.
</p>
<p>
The records themselves have the following general layout:
</p>
<source><![CDATA[
<record id="0x1032" name="Frame" package="org.apache.poi.hssf.record">
<description>The frame record indicates whether there is a border
around the displayed text of a chart.</description>
<author>Glen Stampoultzis (glens at apache.org)</author>
<fields>
<field type="int" size="2" name="border type">
<const name="regular" value="0" description="regular rectangle or no border"/>
<const name="shadow" value="1" description="rectangle with shadow"/>
</field>
<field type="int" size="2" name="options">
<bit number="0" name="auto size"
description="excel calculates the size automatically if true"/>
<bit number="1" name="auto position"
description="excel calculates the position automatically"/>
</field>
</fields>
</record>
]]></source>
<p>
Currently the type can be of type int, float or string. The 'int'
type covers bytes, shorts and integers which is selected using a
size of 1, 2 or 4. An additional type called varword is used to
represent a array of word values where the first short is the length
of the array. The string type generation is only partially
implemented. If choosing string you must select a size of 'var'.
</p>
<p>
The Java records are regenerated each time the record generator is
run, however the test stubs are only created if the test stub does
not already exist. What this means is that you may change test
stubs but not the generated records.
</p>
</s2>
<s2 title="How it Works">
<p>
The record generation works by taking an XML file and styling it
using XLST. Given that XSLT is a little limited in some ways it was
necessary to add a little Java code to the mix.
</p>
<p>
See record.xsl, record_test.xsl, FieldIterator.java,
RecordUtil.java, RecordGenerator.java
</p>
</s2>
<s2 title="Limitations">
<p>
The record generator does not handle all possible record types and
is not ment to. Sometimes it's going to make more sense to generate
the records manually. The main point of this thing is to make the
easy stuff simple.
</p>
<p>
Currently the record generator is optimized to create Excel records.
It could be adapted to create Word records with a little poking
around.
</p>
<p>
Currently the the XSL file that generates the record calls out to
Java objects. This would have been better done as Javascript inside
the XSL file itself. The Java code for the record generation is
currently quite messy with minimal comments.
</p>
</s2>
</s1>
</body>
</document>