Vlado Keselj's home page


Starfish is a system for Perl-based preprocessing and text-embedded programming. This is the Web documentation page generate from the standard Text::Starfish module documentation. The main code of the system is in Starfish.pm. You can find more files in the public project directory. The module is available on CPAN and github as well.


Text::Starfish.pm and starfish - Perl-based System for Preprocessing and Text-Embedded Programming


starfish [ -o=outputfile ] [ -e=initialcode ] [ -replace ] [ -mode=mode ] files...

where files usually contain some Perl code, delimited by <? and !>. Use function echo to produce output to be inserted into the file.


Starfish is a system for Perl-based preprocessing and text-embedded programming, based on a universal approach applicable to many different text styles. You can read the documentation contained in the file report.pdf for an introduction. For an initial understanding about how Starfish works, you can think of Perl code being inserted in arbitary text between <? and !> delimiters, which can be executed in a similar way as PHP code in an HTML file. Some similar projects exist and some of them are listed in "SEE ALSO". Starfish has been unique in several ways. One important difference between starfish and similar programs (e.g., PHP) is that the output does not necessarily replace the code, but it is appended to the code by default.

The package contains two main files: a module file (Starfish.pm) and a small script (starfish) that provides a command-line interface to the module. The options for the script are described in subsection ""starfish_cmd list of file names and options"".


A simple example

Let us have a plain file named example.txt with the following content:

     <? echo "Hello world!" !>

In the command line, run the command:

     starfish example.txt

If we open the file example.txt, the content will be:

     <? echo "Hello world!" !>#+
     Hello world!#-

The same effect would be obtained with the code $O = "Hello world!". This way of updating the file is called the "update" mode of Starfish and it is the default mode. The "replace" mode can be used, but then we should have a different output file, as in the following command:

  starfish -replace -o=example-out.txt example.txt

and the content of the file example-out.txt would now be:

  Hello world!

The module parameters can be changed, and their default values vary according to the text style. THese parameters are described in the description of the set_style method.

HTML Examples

Example 1

If we have an HTML file, e.g., 7.html with the following content:

  <!--<? $O="This code should be replaced by this." !>-->

then after running the command

  starfish -replace -o=7out.html 7.html

the file 7out.html will contain:

  This code should be replaced by this.

The same effect would be obtained with the following line:

  <!--<? echo "This code should be replaced by this." !>-->

Output file permissions

The permissions of the output file will not be changed. But if it does not exist, then:

  starfish -replace -o=7out.html -mode=0644 7.html

makes sure it has all-readable permission.

Example 2

Input file 21.html:

  <!--<? use CGI qw/:standard/;
         echo comment('AUTOMATICALLY GENERATED - DO NOT EDIT');
  <TITLE>Some title</TITLE>
  <!--<? echo "Put this." !>-->


  <TITLE>Some title</TITLE>
  Put this.

Including full links in the anchor tags (<a>) makes the HTML source more difficult to edit and maintain. For example, if we want to attach a link to the anchor text "FCS grad forms", we may want to write it in the following way in HTML:

  <a target="_blank"
  >FCS grad forms</a>

Instead of that, we can in the Starfish file include the following code:

  add_hook('string','<a>FCS grad forms</a>',
    '<a target="_blank" href="'.
    '">FCS grad forms</a>');

and later in the rest of the file simply use:

  <a>FCS grad forms</a>

which will always be replaced with the proper syntax for including the link.

If we use the same anchor text several times, it will be replaced each time with the proper link. The links can be kept at one place, making them easy to maintain and update. We can also generate links is some programmable way if it is appropriate.

Example from a Makefile

  LIST=first second third\
   fourth fifth

  <? echo join "\n", getmakefilelist $Star->{INFILE}, 'LIST', "\n" !>#+

Beside $O, $Star is another predefined variable: It refers to the Starfish object currently processing the text.

TeX and LaTeX Examples

Simple TeX or LaTeX Example

Generating text with a variable replacement:

  %<?echo "
  % When we split the probability reserved for unseen characters equally
  % among the remaining $UnseenNum characters, we obtain the final estimated
  % probabilities:

Example from a TeX file

 % <? $Star->Style('TeX') !>

 % For version 1 of a document
 % <? #add_hook('be',"\n%Begin1","\n%End1",'s/\n%+/\n/g');
 %    #add_hook('be',"\n%Begin2","\n%End2",'s/\n%*/\n%/g');
 %    #For version 2
 %    add_hook('be',"\n%Begin1","\n%End1",'s/\n%*/\n%/g');
 %    add_hook('be',"\n%Begin2","\n%End2",'s/\n%+/\n/g');
 % !>

 %Document 1

 Document 2

LaTeX Example with Final Routine used for Slides

  % -*- compile-command: "make 01s 01"; -*-
  %<? ##read_starfish_conf();
  %  $TexTarget = 'slides';
  %  sfish_add_tag('sl,l', 'echo');
  %  sfish_add_tag('slide', 'echo');
  %  sfish_ignore_outer;
  %  $Star->add_final( sub {
  %    my $r = shift;
  %    $r =~ s/^% -\*- compile-command.*\n//;
  %    $r.= "\\end{document}\n";
  %    return $r;
  %  } );
  % !>

  \section{Course Introduction}

  Not in slide.

  %slide:In slide.

  In slides and lectures.

Example with Test/Release versions (Java)

Suppose you have a stanalone java file p.java, and you want to have two versions:

  p_t.java -- for complete code with all kinds of testing code, and
  p.java -- clean release version.


Copy p.java to p_t.java and modify p_t.java to be like:

  /** Some Java file.  */

  //<? $O = defined($Release) ?
  // "public class p {\n" :
  // "public class p_t {\n";
  public class p_t {

    public static int main(String[] args) {

      //<? $O = "    ".(defined $Release ?
      //qq[System.out.println("Test version");] :
      //qq[System.out.println("Release version");]);
      System.out.println("Release version");//-

      return 0;

In Makefile, add lines for updating p_t.java, and generating p.java (readonly, so that you do not modify it accidentally):

  p.java: p_t.java
        starfish -o=$@ -e='$$Release=1' -mode=0400 $<
  tmp.ind: p_t.java
        starfish $<
        touch tmp.ind

Command-line Examples

The following are the reference examples. For further information, please lookup the explanations of the command-line options and arguments.

starfish -mode=0400 -replace -o=paper.tex -mode=0400 paper.tex.sfish

In the above line, Starfish is used on top of a TeX/LaTeX file. The Starfish is separated from the .tex file to keep the source clean. However, a user in this situation may by mistake start editing the paper.tex file, so we set the output file mode to 0400 to prevent this accidental editing.


Note: This is a quite old part of Starfish and needs a revision. Macros are a form of code folding (related terms: holophrasting, ellusion(?)), expressed in the Starfish framework.

Starfish includes a set of macro features in an experimental phase. There are two modes, hidden macros and not hidden, which are indicated using variable $Star->{HideMacros}, e.g.:

  starfish -e='$Star->{HideMacros}=1' *.sfish
  starfish *.sfish

Macros are activated with:

  <? $Star->defineMacros() !>

In Java mode, a macro can be defined in this way:

  //m!define macro name

After //m!end, a newline is mandatory. After running Starfish, the definition will disapear in this place and it will be appended as an auxdefine at the end of file.

In the following way, it can be defined and expanded in the same place:

  //m!defe macro name

A macro is expanded by:

  //m!expand macro name

When macro is expanded it looks like this:

  //m!expanded macro name

Macro is expanded even in hidden mode by:

  //m!fexpand macro name

and then it is expanded into:

  //m!fexpanded macro name

Hidden macros are put at the end of file in this way:

  //auxdefine macro name

Old macro definition can be overriden by:

  //m!newdefe macro name



After executing a snippet, the contents of this variable represent the snippet output.


More precisely, it is $::Star. $Star is the Starfish object executing the current code snipet (this). There can be a more such objects active at a time, due to executing Starfish from a starfish snippet. The name is introduced into the main namespace, which might be a questionable decision.


If defined, it should be an array of CODE references, which are applied as functions on the final output before writing it out. These are used as final routines, typically to add or remove some of the first lines or finals lines. Each function takes input as a parameter and returns it after processing. The variable should accessed using the method add_final.


Name of the current input file.


Controls the number of iterations. The default value is 1, but we may want to repeat starfishing the text several times, or even until a fix-point is reached. For example, by setting the number of Loops to be at least 2, as in:

    $Star->{Loops} = 2 if $Star->{Loops}<2;

we require Starfish to proces the input in at least two iterations.


Output content of the current processing unit. For example, to use #-style line comments in the replace Starfish mode, one can make a final substitution in an HTML file:

 <!--<? $Star->{Out} =~ s/^#.*\n//mg; !>-->

It is important to have in mind that the contents of this variable is the output processed so far, so any final output processing should be done in a snippet where no new output is produced.


If option -o=* is used, then this variable contains the name of the specified output file.



The method for creation of a new Starfish object. If we are already processing within a Starfish object, we may use a shorter variant $Star->new().

The options, given as arguments, are a list of strings, which may include the following:

-infile=* Specifies the name of the input file (field INFILE). The file will not be read.

-copyhooks Copies hooks from the Star object ($::Star). This option is also available in loadinclude, getinclude, and include, from which it is passed to new. It causes the new object to have similar properties as the current Star object. It could be generalized to include any specified object, or to use the prototype object that is given to the constructor, but there does not seem to be need for this generalization. More precisely, -copyhooks copies the fields: Style, CodePreparation, LineComment, IgnoreOuter, and per-component copies the array hook.


Adds the function referred to by $func_ref to the list of functions to be executed on the output at the end of processing. See also the parameter $Star->{Final}.

$o->add_tag($tag, $action)

Normally used by sfish_add_tag by translating the call to $Star->add_tag($tag, $action). Examples:

  $Star->add_tag('slide', 'ignore');
  $Star->add_tag('slide', 'echo');

See sfish_add_tag for a few more details.

$o->add_hook($ht,...) -- (and function add_hook)

Adds a new hook. The first argument is the hook type, which is a string. If it is used as a function, it will run on the $::Star object. The following is the list of hook types with descriptions:

string, somestring, replacementstring

A simple hook to replace a string with another string. In the update mode, we must take care that the string to be replaced is commented out if needed. For example, after the following embedded code:

  add_hook('string', '<code>App::Utils</code>',
    '<a href="https://metacpan.org/pod/App-Utils" target="_blank">'.
    '<code>App::Utils</code></a>'); ?>

any occurence of <code>App::Utils</code> is replaced with:

  <a href="https://metacpan.org/pod/App-Utils" target="_blank">
be, prefix, suffix

Adding a hook with new prefix (begin delimiter) and suffix (end delimiter). The following example replaces the default hook <?...!> with a new one <?new ...!>:

  rm_hook('be', '<?', '!'.'>'); # remove default hook (notice that we avoid
                                # literal ending delimiter '!>' in order
                                # not to be confused with default suffix
  add_hook('be', '<?new ', '!'.'>'); # adding a new hook
regex, regex, replace

The hook type regex is followed by a regular expression and a replace argument. Whenever a regular expression is matched in text, it is ``starfished'' according to the argument replace. If the argument replace is the string ``comment'', it is treated as the comment. If the argument replace is code, it is used as the evaluation code. For example, the following source in an HTML file:

  <!--<? $Star->add_hook('regex', qr/^.section:(\w+)\s+(.*)/,
  sub { $_="<a name\"$_[2]\"><h3>$_[3]</h3</a>" }) !>-->

  line before
  .section:overview Document Overview
  line after

will produce the following output, in the replace mode:

  line before
  <a name"overview"><h3>Document Overview</h3</a>
  line after
ht:re2, regex, replace

The hook type ht:re2 is a special type used for Python and Makefile styles in order to capture indentation, which needs to be maintained in the output. It is regular expression based.

$o->addHook -- deprecated, should use add_hook

This method is deprecated. It will be gradually replaced with add_hook, which is better defined since it includes hook type.

Adds a new hook. The method can take two or three parameters:

 ($prefix, $suffix, $evaluator)


 ($regex, $replacement)

In the case of three parameters ($prefix, $suffix, $evaluator), the parameter $prefix is the starting delimiter, $suffix is the ending delimiter, and $evaluator is the evaluator. The parameters $prefix and $suffix can either be strings, which are matched exactly, or regular expressions. An empty ending delimiter will match the end of input. The evaluator can be provided in the following ways:

special string 'default'

in which case the default Starfish evaluator is used,

special strings 'ignore' and 'echo'

'ignore' ignores the hook and produces no echo, 'echo' simply echos the contests between the delimiters.

other strings

are interpreted as code which is embedded in an evaluator by providing a local $_, $self which is the current Starfish object, $p - the prefix, and $s the suffix. After executing the code $p.$_.$s is returned, unless in the replacement mode, in which $_ is returned.

code reference (sub {...})

is interpreted as code which is embedded in an evaluator. The local $_ provides the captured string. Three arguments are also provided to the code: $p - the prefix, $_, and $s - the suffix. The result is the value of $_.

For the format with two parameters, ($regex, $replacement), currently in this mode addHook understands replacement 'comment' and code reference (e.g., sub { ... }). The replacement 'comment' will repeat the token in the non-replace mode, and remove it in the replace mode; e.i., equivalent to no echo. The regular expression is matched in the multi-line mode, so ^ and $ can be used to match beginning and ending of a line. (Caveat: Due to the way how scanner works, beginning of a line starts after the end of previously matched token.)


 $Star->addHook(qr/^#.*\n/, 'comment');


Sets the mode for ignoring the outer text in the replace mode. The function sfish_ignore_outer does the same on the default object Star. If an argument is given, it is used to set the mode, so as a consequence the mode can be turned off by giving the argument ''.


Or just last_update(), returns the date of the last update of the output.


Similar to the function starfish_cmd, but it expects already built Starfish object with properly set options. Actually, starfish_cmd calls this method after creating the object and returns the object.

$o->rmHook($p,$s) -- deprecated, should use rm_hook

Removes a hook specified by the starting delimiter $p, and the ending delimiter $s.

$o->rm_hook($ht,...) -- and function rm_hook

Removes a hook. Example:

 rm_hook('be', '<?', '!>');  # removes all hooks with give begin and end


Removes all hooks. If no hooks are added, then after exiting the current snippet it will not be possible to detect another snippet later. A typical usage could be as follows:

    $Star->add_hook('be', '<?starfish ','?>', 'default');

$o->setStyle($s) -- deprecated, shoud use set_style

Deprecated method. The method or function set_style should be used.

set_style method or function

Sets a particular style of the source file. If used as function, the object $::Star is used as the "self" object. Currently implemented options are: html, java, makefile, perl, ps, python, and tex (same as latex, TeX). If the parameter $s is not given, the stile given in $o-{STYLE}> will be used if defined, otherwise it will be guessed from the file name in $o-{INFILE}>. If it cannot be correctly guessed, it will be the Perl style.

Setting a style can have various side effects, but it typically involves setting the following variables:

 $o->{Style}            # style string id
 $o->{CodePreparation}  # function to clean the code before running
 $o->{LineComment}      # string starting a line comment
 $o->{OutDelimiters}    # array ref with four elements: $b1, $b2 for
                        # starting output delimiter, and $e1, $e2 for
                        # the ending output delimiter; $b1 and $e1
                        # must not end with a digit, and $b2 and $e2
                        # must not start with a digit
 $o->{IgnoreOuter}      # boolean variable to ignore outer text, false
                        # by default
 $o->{hook}             # array ref, list of hooks


include( filename and options ) -- starfish a file and echo

Reads, starfishes the file specified by file name, and echos the contents. Similar to PHP include. Uses getinclude function.

getinclude( filename and options ) -- starfish a file and return

Reads, starfishes the file specified by file name, and returns the contents (see also include to echo the content implicitly). By default, the program will not break if the file does not exist. The option -noreplace will starfish file in a non-replace mode. The default mode is replace and that is usually the mode that is needed in includes (non-replace may lead to a suprising behaviour). The option -require will cause program to croak if the file does not exist. It is similar to the PHP function require. A special function named require is not used since require is a Perl reserved word. Another interesting option is -copyhooks, for using hooks and some other relevant properties from the Star object ($::Star). This option is eventually passed to new, so you can see the constructor new for more details.

The code for get include is the following:

 sub getinclude($@) {
     my $sf = loadinclude(@_);
     return $sf->{Out};

and it can be used as a useful template for using loadinclude directly. The function loadinclude creates a Starfish object, and reads the file, however it is not digested yet, so one can modify the object before this.

loadinclude( filename and options ) -- load file and get ready to digest

The first argument is a filename. Loadinclude will interpret the options -replace, -noreplace, and -require. A Starfish object is created by passing the file name as an -infile argument, and by passing other options as arguments. The file is read and the object is returned. By default, the program will not break if the file does not exist or is not readable, but it will return undef value instead of an object. See also documentation about include, getinclude, and new.

-noreplace option will set up the Starfish object in the no-replace mode. The default mode is replace and that is usually the mode that is needed in includes. The option -require will cause program to croak if the file does not exist. An interesting option is -copyhooks, which is documented in the new method.


This function is usually called at the begining of a starfish file, in order to read local configuration. it tests whethere there exists a filed named starfish.conf in the current directory. If it does exist, it checks for the same file in the parent directory, then gran-parent directory, etc. Once the process stops, is starts executing the configuration files in the order from first ancestor down. For each file, it changes directory to the corresponding directory, and requires (in Perl style) the file in the package main.

sfish_add_tag ( tag, action )

Used to introduce simple tags such as line tag %sl,l: and %<sl,l>...</sl,l> in TeX/LaTeX for inclusion and exclusion of text. Example:

     sfish_add_tag('sl,l', 'echo');
     sfish_add_tag('slide', 'ignore');

and, for example, the following text is included:

     %sl,l:some text to the end of line
     more lines of text

and the following text is excluded:

     %slide:this line is excluded
     more lines of text excluded


Sets the default object $Star in the mode for ignoring outer text if in the replace mode. If an argument is given, it is used to set the mode, so as a consequence the mode can be turned off with sfish_ignore_outer('').

starfish_cmd list of file names and options

The function starfish_cmd is called by the script starfish with the @ARGV list as the list of arguments. The function can also be used from Perl code to "starfish" a file, e.g.,

    starfish_cmd('somefile.txt', '-o=outfile', '-replace');

The arguments of the functions are provided in a similar fashion as argument to the command line. As a reminder, the command usage of the script starfish is:

starfish [ -o=outputfile ] [ -e=initialcode ] [ -replace ] [ -mode=mode ] file...

The options are described below:


specifies an output file. By default, the input file is used as the output file. If the specified output file is '-', then the output is produced to the standard output.


specifies the initial Perl code to be executed.


will cause the embedded code to be replaced with the output. WARNING: Normally used only with -o.


specifies the mode for the output file. By default, the mode of the source file is used (the first one if more outputs are accumulated using -o). If an output file is specified, and the mode is specified, then starfish will set temporarily the u+w mode of the output file in order to write to that file, if needed.

Those were the options.

echo list

appends all elements of the list to the special variable $0.



The function reads strings and translates it into an array of records according to DB822 (db8 for short) data format. If the string starts with 'file=' then the rest of the string is treated as a file name, which contents replaces the string in further processing. The string is translated into a list of records (hashes) and a reference to the list is returned. The records are separated by empty line, and in each line an attribute and its value are separated by the first colon (:). A line can be continued using backslash (\) at the end of line, or by starting the next line with a space or tab. Ending a line with \ will replace "\\\n" with "\n" in the string, otherwise "\n[ \t]" are kept as they are. Lines starting with the hash sign (#) are considered comments and they are ignored, unless they are part of a multi-line string. An example is:

  name: J. Public
  phone: 000-111


If an attribute is repeated, it will be renamed to an attribute of the form att-1, att-2, etc.



returns the current year in string format.


Returns modification time of this file (in format of Perl time).


Returns modification date of this file (in format: Month DD, YYYY).


appendfile $filename, @list

appends list elements to the file.

getfile $filename

reads the contents of the file into a string or a list.

getmakefilelist($makefilename, $var)

returns a list, which is a list of words assigned to the variable $var in the makefile named $makefilename; for example:

  FILE_LIST=file1 file2 file3\

  <? echo join "\n", getmakefilelist $Star->{INFILE}, 'FILE_LIST' !>

Embedded variables are not handled.

putfile $filename, @list

Opens the file $filename, wries the list elements to the file, and closes it. `putfile filename' will only touch the file.


There is a set of predefined styles for different input files: HTML (html), HTML templating style (html.sfish), TeX (tex), Java (java), Makefile (makefile), PostScript (ps), Python (python), and Perl (perl).

HTML Style (html)

HTML Templating Style (html.sfish)

This style is similar to the HTML style, but it is supposed to be run in the replace mode towards a target .html file, so it allows for more hooks. The character # (hash) at the beginning of a line denotes a comment.

Makefile Style (makefile)

The main code hooks are <? and >.

Interestingly, the makefile style has similar special requirements as Python. For example, in the following expansion:

 starfish: tmp
         starfish Makefile
         #<? if (-e "file.tex.sfish")
         #{ echo "\tstarfish -o=tmp/file.tex -replace file.tex.sfish\n" } !>#+
         starfish -o=tmp/file.tex -replace file.tex.sfish

it is convenient to have the embedded output indented in the same way as the embedded code.


get_verbatim_file( filename )

Specific to LaTeX mode. Reads textual file filename and returns a string ready for inclusion in a LaTeX document. It untabifies the file contests for proper representation of whitespace. The function code is basically:

    return "\\begin{verbatim}\n".
           "\\ end{verbatim}\n";

Note: There is no space betwen \\ and end{verbatim}.

htmlquote( string )

The following definition is taken from the CIPP project.

(http://aspn.activestate.com/ASPN/CodeDoc/CIPP/CIPP/Manual.html, link does not seem to be active any more)

This command quotes the content of a variable, so that it can be used inside a HTML option or <TEXTAREA> block without the danger of syntax clashes. The following conversions are done in this order:

       &  =>  &amp;
       <  =>  &lt;
       "  =>  &quot;


The script swallows the whole input file at once, so it may not work on small-memory machines and with huge files.


I'd like to thank Steve Yeago, Tony Cox, Tony Abou-Assaleh for comments, Charles Ikeson for suggesting the include function and other comments, and Mohammad S Anwar for corrections in Perl packaging.


 2001-2021 Vlado Keselj https://web.cs.dal.ca/~vlado
           and contributing authors:
      2007 Charles Ikeson (overhaul of test.pl)

This script is provided "as is" without expressed or implied warranty. This is free software; you can redistribute it and/or modify it under the same terms as Perl itself.

The latest version can be found at http://web.cs.dal.ca/~vlado/srcperl/.


There are several projects similar to Starfish. Some of them are text-embedded programming projects such as PHP with different programming languages, and there are similar Perl-based projects. When I was thinking about a need of a framework like this one (1998), I have found ePerl project. However, it was too heavy weight for my purposes, and it did not support the "update" mode, vs. replace mode of operation. I learned about more projects over time and they are included in the list below.

[ePerl] ePerl

This script is somewhat similar to ePerl, about which you can read at

http://www.ossp.org/pkg/tool/eperl/. It was developed by Ralf S. Engelshall in the period from 1996 to 1998.



[ePerl-h] ePerl hack by David Ljung Madison

This is a Perl script simulating the ePerl functionality, but with obviously much lower weight. It is developed by David Ljung Madison, and can be found at the URL: http://marginalhacks.com/Hacks/ePerl/

[Text::Template] Perl module Text::Template by Mark Jason Dominus.

http://search.cpan.org/~mjd/Text-Template/ Text::Template is a module with similar functionality as Starfish. An interesting similarity is that the output variable in Text::Template is called $OUT, compared to $O in Starfish.

[HTML::Mason] Perl module HTML::Mason by Jonathan Swartz, Dave Rolsky, and Ken Williams.

http://search.cpan.org/~drolsky/HTML-Mason-1.28/lib/HTML/Mason/Devel.pod The module HTML::Mason can also be seen as an embedded Perl system, but it is a larger system with the design objective being a "high-performance, dynamic web site authoring system".

[HTML::EP] Perl Module HTML::EP - a system for embedding Perl into HTML, by Jochen Wiedmann.

http://search.cpan.org/~jwied/HTML-EP-MSWin32/lib/HTML/EP.pod It seems that the module was developed in 1998-99. Provides a good CGI support, run-time support, session handling, a database server interface.