perldl - Man Page

The PDL reference manual and online help are available from within  perldl, using the help and apropos commands (which may also be abbreviated ? and ??.)   The help command alone prints a summary of help syntax, and help <module-name> will print POD documentation from the module you mention (POD is the Perl format for embedding  documentation in your perl code; see perlpod for details).

If you include POD documentation in your autoload subroutines (see path mechanism below), then both help and apropos will find it  and be able to format and display it on demand.

If you have the perl modules ReadLines and ReadKeys installed, then perldl supports a history and line-editing mechanism using editing keys similar to emacs. The last 500 commands are always stored in the file .perldl_hist in your home directory between sessions. Set $PERLDL::HISTFILESIZE to change the number of lines saved. The command l [number] shows you the last number commands you typed where number defaults to 20.

e.g.:

   bash$ perldl
   ReadLines enabled
   pdl> $x = rfits "foo.fits"
   BITPIX =  -32  size = 88504 pixels
   Reading  354016 bytes
   BSCALE =  &&  BZERO =

   pdl> imag log($x+400)
   Displaying 299 x 296 image from 4.6939525604248 to 9.67116928100586 ...

If you enter a simple command at the perldl command line, it is immediately executed in a Perl eval().  The environment is almost identical to that within a perl script, with some important exceptions:

• $_ is not preserved across lines

  $_ is used to hold the command line for initial processing, so at the beginning of processing of each command line, $_ contains the command itself. Use variables other than $_ to store values across lines.

• Scope is not preserved across lines

  Each command line is executed in a separate eval block within perl, so scoping commands such as my and local may not perform exactly as expected -- in particular, if you declare a variable with my, it is local to the particular command line on which you typed the my command, which means that it will evaporate before the next prompt is printed. (You can use my variables in a multi-line block or to isolate values within  a single command line, of course).

    NOTE: pdl2 preserves lexical scope between lines.

• Execution is immediate

  Under most circumstances, as soon as you end a line of input the line is parsed and executed.  This breaks Perl's normal dependence on semicolons as command delimiters.  For example, the two-line expression

    print "Hello ",
       "world";

  prints the phrase Hello world in Perl, but (under most circumstances)  Hello in perldl.

• Multi-line execution

  In multiline mode (which is enabled by default, see Shell variables, below), perldl searches for searches for block-like constructs with curly braces, parentheses, quotes, and related delimiters.  If you leave such a construct open, perldl accepts more lines of input until you close the construct or explicitly end the multi-line expression with ^D.   Following the example above, the phrase

    { print "Hello ",
         "world"; }

  will print "Hello world" from either Perl or (in multi-line mode) perldl.

  Warning: The multi-line parsing uses Damian Conway's Text::Balanced module, which contains some flaws -- so it can be fooled by quote-like operators such as q/.../, included POD documentation, multi-line << quotes, and some particularly bizarre-but-valid m/.../ matches and s/.../.../ substitutions.  In such cases, use ^D to close out the multi-line construct and force compilation-and-execution.

If you want to preserve this behavior in a script (for example to replay a command journal file; see below on how to create one), you can use pdl instead of perl as the interpreter in the script's initial shebang line.

A perldl session can be terminated with any of the commands quit, exit or the shorthands x or q.  If EOF handling is switched on (the default) you can also type ^D at the command prompt.

If the command input is NOT a terminal (for example if you are running from a command journal file), then EOF will always terminate perldl.

Commands executed within perldl can be terminated prematurely using Ctrl-C (or whichever key sequence sends an INT signal to the process on your terminal). Provided your PDL code does not ignore sigints this should throw you back at the perldl command prompt:

  pdl> $result = start_lengthy_computation()
   <Ctrl-C>
  Ctrl-C detected

  pdl>

As of 2.077, this requires Sys::SigAction to be installed (without that, and before 2.077, for Perl >5.8 it didn't actually interrupt things).

• The shell aliases p to be a convenient short form of print, e.g.

     pdl> p ones 5,3
  
     [
      [1 1 1 1 1]
      [1 1 1 1 1]
      [1 1 1 1 1]
     ]

• q and x are short-hand for quit.

• l lists the history buffer

    pdl> l # list last 20 commands
  
    pdl> l 40 # list last 40 commands

• ? is an alias for help

    pdl> ? pdl2    # get help for new pdl2 shell

• ?? is an alias for apropos

    pdl> ?? PDL::Doc

• help, apropos, usage and sig: all words after these commands are used verbatim and not evaluated by perl. So you can write, e.g.,

    pdl> help help

  instead of

    pdl> help 'help'

• with_time runs the following code-block, and tells you how long it took, in milliseconds. Requires Time::HiRes.

    pdl> with_time { print +($A->matmult($B))->info, "\n" } for 1..5;

• gv uses "pdumpgraph" in PDL::Core etc to visualise the given ndarray, with all connected transformations, etc, into the given file. Requires Graph and GraphViz2.

    pdl> gv $A, 'output.png'

perldl and pdl support several command-line options to adjust the behavior of the session.  Most of them are equivalent to commands that can be entered at the pdl> prompt.  They are:

-glut

Load OpenGL when starting the shell (the perl OpenGL module, which is available from CPAN must be installed).  This enables readline event loop processing.  Don't use with -tk.

-tk

Load Tk when starting the shell (the perl Tk module, which is available from CPAN must be installed).  This enables readline event loop processing.  Don't use with -glut.

-f file

Loads the file before processing any user input. Any errors during the execution of the file are fatal.

-w

Runs with warning messages (i.e. the normal perl -w warnings) turned-on.

-M module

Loads the module before processing any user input. Compare corresponding perl switch.

-m module

Unloads the module before processing any user input.

-I directory

Adds directory to the include path. (i.e. the @INC array) Compare corresponding perl switch.

-V

Prints a summary of PDL config. This information should be included with any PDL bug report. Compare corresponding perl switch.

If the file ~/.perldlrc is found it is sourced at start-up to load default modules, set shell variables, etc. If it is NOT found the distribution file PDL/default.perldlrc is read instead. This loads various modules considered useful by default, and which ensure compatibility with v1.11. If you don't like this and want a more streamlined set of your own favourite modules simple create your own ~/.perldlrc.  You may wish to start from the existing PDL/default.perldlrc as a template since it will not be sourced once you replace it with your own version.

To set even more local defaults the file  local.perldlrc (in the current directory) is sourced if found. This lets you load modules and define subroutines for the project in the current directory.

The name is chosen specifically because it was found hidden files were NOT wanted in these circumstances.

The startup file should normally include "use PDL::AutoLoader;", as  many of the nicer interactive features won't work without it.

Shell variables: (Note: if you don't like the defaults change them in ~/.perldlrc)

• $PERLDL::ESCAPE  - default value '#'

  Any line starting with this character is treated as a shell escape. The default value is chosen because it escapes the code from the standard perl interpreter.

• $PERLDL::HISTFILESIZE  - default value 500

  This is the number of lines of perldl shell command history to keep.

• $PERLDL::PAGER - default value more

  External program to filter the output of commands.  Using more prints output one screenful at a time.  On Unix, setting page(1) and $PERLDL::PAGER to tee -a outfile will keep a record of the output generated by subsequent perldl commands (without paging).

• $PERLDL::PROMPT - default value 'pdl> '

  Enough said  But can also be set to a subroutine reference, e.g. $PERLDL::PROMPT = sub {join(':',(gmtime)[2,1,0]).'> '} puts the current time into the prompt.

• $PERLDL::MULTI - default value 1

  If this is set to a true value, then perldl will parse multi-line  perl blocks: your input will not be executed until you finish a line with no outstanding group operators (such as quotes, blocks, parenthesis, or  brackets) still active.  Continuation lines have a different prompt that shows you what delimiters are still active.

  Note that this is not (yet!) a complete perl parser.  In particular, Text::Balanced appears to be able to ignore quoting operatores  like q/ ... / within a line, but not to be able to extend them across lines.  Likewise, there is no support for the '<<' operator.

  Multiline conventional strings and {}, [], and () groupings are well supported.

• $PERLDL::NO_EOF - default value 0 / 1 on MSWin32

  Protects against accidental use of "^D" from the terminal.  If this is set to a true value, then you can't accidentally exit perldl by typing "^D".  If you set it to a value larger than 1 (and PERLDL::MULTI is set), then you can't use "^D" to exit multiline commands either.  If you're piping commands in from a file or pipe, this variable has no effect.

• $HOME

  The user's home directory

• $PERLDL::TERM

  This is the Term::ReadLine object associated with the perldl shell. It can be used by routines called from perldl if your command is interactive.

• $PDL::toolongtoprint

  The maximal size pdls to print (defaults to 10,000 elements). This is not just a perldl or pdl2 variable but it is something that is usually needed in an interactive debugging session.

A useful idiom for developing perldl scripts or editing functions on-line is

  pdl> # emacs script &
    -- add perldl code to script and save the file
  
  pdl> do 'script'

-- substitute your favourite window-based editor for 'emacs' (you may also need to change the '&' on non-Unix systems).

Running "do 'script'" again updates any variables and function definitions from the current version of 'script'.

PDL scripts are just perl scripts that happen to use PDL (and possibly PDL::NiceSlice).  But for the truly lazy, perldl can be invokes as a script interpreter.  Because perldl is itself an interpreted perl script, most unices won't allow you to say "#!/usr/bin/perldl"  at the top of your script.

Instead, say "#!/usr/bin/pdl" and your script will be executed exactly as if you typed it, line-by-line, into the perldl shell.

NOTE: This feature is used by default by PDL::NiceSlice. See below for more about slicing at the perldl prompt

In some cases, it is convenient to process commands before they are sent to perl for execution. For example, this is the case where the shell is being presented to people unfamiliar with perl but who wish to take advantage of commands added locally (eg by automatically  quoting arguments to certain commands).

*NOTE*: The preprocessing interface has changed from earlier versions! The old way using $PERLDL::PREPROCESS will still work but is strongly deprecated and might go away in the future.

You can enable preprocessing by registering a filter with the preproc_add function. preproc_add takes one argument which is the filter to be installed. A filter is a Perl code reference (usually set in a local configuration file) that will be called, with the current command string as argument, just prior to the string being executed by the shell. The modified string should be returned. Note that you can make perldl completely unusable if you fail to return the modified string; quitting is then your only option.

Filters can be removed from the preprocessing pipeline by calling preproc_del with the filter to be removed as argument.  To find out if a filter is currently installed in the preprocessing pipeline use preproc_registered:

  pdl> preproc_add $myfilter unless preproc_registered $myfilter;

Previous versions of perldl used the variable $PERLDL::PREPROCESS. This will still work but should be avoided. Please change your scripts to use the preproc_add etc functions.

The following code would check for a call to function 'mysub' and bracket arguments with qw.

  $filter = preproc_add sub {
     my $str = shift;
     $str =~ s/^\s+//;  # Strip leading space
     if ($str =~ /^mysub/) {
        my ($command, $arguments) = split(/\s+/,$str, 2);
        $str = "$command qw( $arguments )" 
        if (defined $arguments && $arguments !~ /^qw/);
     };
     # Return the input string, modified as required
     return $str;
   };

This would convert:

  pdl> mysub arg1 arg2

to

  pdl> mysub qw( arg1 arg2 )

which Perl will understand as a list.  Obviously, a little more effort is required to check for cases where the caller has supplied a normal list (and so does not require automatic quoting) or variable interpolation is required.

You can remove this preprocessor using the preproc_del function which takes one argument (the filter to be removed, it must be the same coderef that was returned from a previous preproc_add call):

  pdl> preproc_del $filter;

An example of actual usage can be found in the perldl script. Look at the function trans to see how the niceslicing preprocessor is enabled/disabled.

PDL::NiceSlice introduces a more convenient slicing syntax for ndarrays. In current versions of perldl and pdl2 niceslicing is enabled by default (if the required CPAN modules are installed on your machine).

At startup perldl will let you know if niceslicing is enabled. The startup message will contain info to this end, something like this:

   perlDL shell v1.XX
    PDL comes with ABSOLUTELY NO WARRANTY. For details, see the file
    'COPYING' in the PDL distribution. This is free software and you
    are welcome to redistribute it under certain conditions, see
    the same file for details.
   ReadLines, NiceSlice  enabled
   Reading /home/csoelle/.perldlrc...
   Type 'demo' for online demos
   Loaded PDL v2.XX

When you get such a message that indicates NiceSlice is enabled you can use the enhanced slicing syntax:

  pdl> $x = sequence 10;
  pdl> p $x(3:8:2)

For details consult PDL::NiceSlice.

PDL::NiceSlice installs a filter in the preprocessing pipeline (see above) to enable the enhanced slicing syntax. You can use a few commands in the perldl shell to switch this preprocessing on or off and also explicitly check the substitutions that the NiceSlice filter makes.

You can switch the PDL::NiceSlice filter on and off by typing

  pdl> trans # switch niceslicing on

and

  pdl> notrans # switch niceslicing off

respectively. The filter is on by default.

To see how your commands are translated switch reporting on:

  pdl> report 1;
  pdl> p $x(3:8:2)
  processed p $x->slice([3,8,2])
  [3 5 7]

Similarly, switch reporting off as needed

  pdl> report 0;
  pdl>  p $x(3:8:2)
  [3 5 7]

Reporting is off by default.

The variable @PERLDL::AUTO is a simple list of perl code strings and/or code reference. It is used to define code to be executed automatically every time the user enters a new line.

A simple example would be to print the time of each command:

  pdl> push @PERLDL::AUTO,'print scalar(gmtime),"\n"'
  
  pdl> print zeroes(3,3)
  Sun May  3 04:49:05 1998
  
  [
   [0 0 0]
   [0 0 0]
   [0 0 0]
  ]
  
  pdl> print "Boo"
  Sun May  3 04:49:18 1998
  Boo
  pdl>

Or to make sure any changes in the file 'local.perldlrc' are always picked up :-

  pdl> push @PERLDL::AUTO,"do 'local.perldlrc'"

This code can of course be put *in* 'local.perldlrc', but be careful :-) [Hint: add unless ($started++) to above to ensure it only gets done once!]

Another example application is as a hook for Autoloaders (e.g. PDL::AutoLoader) to add code too which allows them to automatically re-scan their files for changes. This is extremely convenient at the interactive command line. Since this hook is only in the shell it imposes no inefficiency on PDL scripts.

Finally note this is a very powerful facility - which means it should be used with caution!