Lembas::Specification -- Technical specification for the Lembas syntax and commands


This is the full specification for the current version of Lembas. For usage documentation, see the main Lembas manual page.


Whitespace is important.


Any line starting at column 0 is interpreted as an internal Lembas command, or a comment. Note that output lines may also have non-space characters at column 0 if they are "tagged".


Comments start with a "#" on the first column of any line.

Lembas uses a shebang-like syntax to specify what program or interpreter ought to be started to run commands or generate output. For example:

  #!/bin/bash -r

starts a restricted shell interpreter. Shebang lines must be the first line of the Lembas script file, or they will be treated as regular comments. Shebangs are overridden by the Lembas constructor's shell argument, if present.


All Lembas commands match qr/\w+/, and may take any number of arguments depending on the command. Lembas supports basic shell-like quoting and quote escaping, through Text::ParseWords' quotewords function.

See also the COMMANDS section of this document, for a list of currently supported commands and their parameters.


Input and output lines start at column 4.


Input lines start with a dollar sign and space. They will be directly passed on to the current interpreter, with an appended newline (the current platform's "\n").

      $ export FOOTMPDIR=$(mktemp -d)
      $ cd $FOOTMPDIR


This is the default matching style. Lines of text output by the interpreter will be matched with eq to lines of literal matching output. For example:

  # note that `cd' has no output
      $ cd /tmp
      $ pwd

Lembas strips any output of its final CRLF (or just LF) and some other things, in order for test files to be reasonably easy to write:

  • carriage returns

  • ANSI escape sequences

  • non-backspace characters followed by a backspace character

This means that you still have a chance of writing a working test file even if the interpreter prints "funny colored lines that don't move":

  Starting foobar daemon...    [WAITING]
  Starting foobar daemon...    [  OK   ]
  foobar daemon started successfully!


Regex matching output is differentiated from literal matching output by the presence of a "re: " tag at column 0. In a regex matching output line, any Perl regex is allowed. The matching test is done under the effects of use re 'eval', so you can do tricky things like

  package LembasWrap {
      # match hex changeset hashes
      our $hg_changeset_re = qr/[a-f0-9]{12}/;
  eval {
      my $lembas = Lembas->new_from_test_spec(...);
  }; ...

and in your Lembas script file:

  # Checking that everything looks good
      $ hg log
  re: changeset:   0:(??{${LembasWrap::hg_changeset_re}})
      tag:         tip
      user:        Fabrice Gabolde <>
  re: date:        .*
      summary:     created repo and added a file

with delayed regex interpolation.

Like in literal matching, output lines are sanitized by removing line endings and color.



  fastforward QUANTIFIER

Skip ahead until the next expected output matches. Does not generate a test.

This command works by trying output lines one by one against the next match (literal or regex). When they fail, the failed match is put back into the command queue, and Test::Builder history is rewritten to pretend everything went fine. Any Test::Builder output during this phase is suppressed.

When an output line finally matches the next test, Lembas exits fastforwarding mode and outputs the results of the successful test as if nothing had happened.

QUANTIFIER has no use for the moment.



This special command can only be used as the very first command in a Lembas script. It kickstarts Lembas into matching output (because Lembas expects a command to generate output, it gets confused when output is generated without a command). You can use it to match e.g. license and copyright information.

Because preamble acts much like a regular input line, it generates a "no output left unmatched" test.


  wait_less_than VALUE UNIT

Start a timer; if some output does not arrive within a certain time, report a failure and bail out from the test script (it is unlikely the rest of the script is going to work anyway). If some output does arrive, report a success and kill the timer.

This uses alarm, in case it is important for your test script to be able to sleep.

VALUE may be any value. UNIT may be "second", "seconds", "minute" or "minutes".



The yield command exits the Lembas loop mid-tests so that you can perform some other tasks. You can resume the Lembas tests by calling the run method again.


Fabrice Gabolde <>


Copyright (C) 2013 by Fabrice Gabolde

This library is free software; you can redistribute it and/or modify it under the same terms as Perl itself, either Perl version 5.10.0 or, at your option, any later version of Perl 5 you may have available.