FileSets are groups of files. You can include or exclude specific files and patterns to/from a FileSet. The use of patterns is explained below. For a start, look at the following example:

<fileset dir="/tmp" id="fileset1">
  <include name="sometemp/file.txt" />
  <include name="othertemp/**" />
  <exclude name="othertemp/file.txt" />
</fileset>

<fileset dir="/home" id="fileset2">
  <include name="foo/**" />
  <include name="bar/**/*.php" />
  <exclude name="foo/tmp/**" />
</fileset>

The use of patterns is quite straightforward: If you simply want to match a part of a filename or dirname, you use *. If you want to include multiple directories and/or files, you use **. This way, filesets provide an easy but powerful way to include files.

FileLists, like FileSets, are collections of files; however, a FileList is an explicitly defined list of files -- and the files don't necessarily have to exist on the filesystem.

Besides being able to refer to nonexistent files, another thing that FileLists allow you to do is specify files in a certain order. Files in FileSets are ordered based on the OS-level directory listing functions, in some cases you may want to specify a list of files to be processed in a certain order -- e.g. when concatenating files using the <append> task.

<filelist dir="base/" files="file1.txt,file2.txt,file3.txt"/>

<!-- OR: -->
<filelist dir="basedir/" listfile="files_to_process.txt"/>

FilterChains can be compared to Unix pipes. Unix pipes add a great deal of flexibility to command line operations; for example, if you wanted to copy just those lines that contained the string blee from the first 10 lines of a file called foo to a file called bar, you could do:

cat foo | head -n10 | grep blee > bar

Something like this is not possible with the tasks and types that we have learned about thus far, and this is where the incredible usefulness of FilterChains becomes apparent. They emulate Unix pipes and provide a powerful dimension of file/stream manipulation for the tasks that support them.

FilterChain usage is quite straightforward: you pass the complex Phing type filterchain to a task that supports FilterChains and add individual filters to the FilterChain. In the course of executing the task, the filters are applied (in the order in which they appear in the XML) to the contents of the files that are being manipulated by your task.

<filterchain>
  <replacetokens>
    <token key="BC_PATH" value="${top.builddir}/"/>
    <token key="BC_PATH_USER" value="${top.builddir}/testsite/user/${lang}/"/>
  </replacetokens>

  <filterreader classname="Phing\Filter\TailFilter">
    <param name="lines" value="10"/>
  </filterreader>
</filterchain>

The code listing above shows you some example of how to use filter chains. For a complete reference see Appendix D, Core Types. This filter chain would replace all occurrences of BC_PATH and BC_PATH_USER with the values assigned to them in lines 4 and 5. Additionally, it will only return the last 10 lines of the files.

Notice above that FilterChain filters have a "shorthand" notation and a long, generic notation. Most filters can be described using both of these forms:

<replacetokens>
  <token key="BC_PATH" value="${top.builddir}/"/>
  <token key="BC_PATH_USER" value="${top.builddir}/testsite/user/${lang}/"/>
</replacetokens>

<!-- OR: -->

<filterreader classname="Phing\Filter\ReplaceTokens">
  <param type="token" name="BC_PATH" value="${top.builddir}/"/>
  <param type="token" name="BC_PATH"
  value="${top.builddir}/testsite/user/${lang}/"/>
</filterreader>

As the pipe concept in Unix, the filter concept is quite complex but powerful. To get a better understanding of different filters and how they can be used, take a look at any of the many uses of FilterChains in the build files for the binarycloud Bibliography project.

With FilterChains and filters provide a powerful tool for changing contents of files, mappers provide a powerful tool for changing the names of files.

To use a Mapper, you must specify a pattern to match on and a replacement pattern that describes how the matched pattern should be transformed. The simplest form is basically no different from the DOS copy command:

copy *.bat *.txt

In Phing this is the glob Mapper:

<mapper type="glob" from="*.bat" to="*.txt"/>

Phing also provides support for more complex mapping using regular expressions:

<mapper type="regexp" from="^(.*)\.conf\.xml$$" to="\1.php"/>

Consider the example below to see how Mappers can be used in a build file. This example includes some of the other concepts introduced in this chapter, such as FilterChains and FileSets. If you don't understand everything, don't worry. The important point is that Mappers are types too, which can be used in tasks that support them.

<copy>
  <fileset dir=".">
    <include name="*.ent.xml"/>
  </fileset>

  <mapper type="regexp" from="^(.*)\.ent\.xml$" to="\1.php"/>

  <filterchain>
    <filterreader classname="Phing\Filter\XsltFilter">
      <param name="style" value="ent2php.xsl"/>
    </filterreader>
  </filterchain>
</copy>

For a complete reference, see Appendix D, Core Types