=head1 NAME
Lingua::PT::Conjugate.pm - Module for Conjugating Portuguese verbs
=head1 DESCRIPTION
This module contains various routines for conjugating Portuguese verbs.
=head1 USAGE
use Lingua::PT::Conjugate;
This module pollutes your namespace with the function C<conjug>,
that takes just the same arguments as the program C<conjug> (which is
just a wrapper).
perl -e 'use Lingua::PT::Conjugate; print conjug("programar","perfeito")'
programar :
perf programei programaste programou programamos programaram
See the C<conjug> manpage for more examples.
=head1 SYNOPSIS
$a = conjug( [$options,] $verb [,$tense ... ] [,$person ... ])
=head2 Options define how the output will be formatted.
=over 4
=item 'q' : 'Quiet', only return the conjugated forms.
=item 'v' : 'Verbose', return the infinitive, the tenses, the
conjugated forms and say if the verb is irregular or follows a model
(Default).
=item 's' : 'Single', returned string is a single line. 'quiet' option
is forced, but can be toggled off by a subsequent 'v' option.
=item 'r' : 'Row', each row represents a person.
=item 'c' : 'Column', each column represents a person (Default).
=item 'o' : 'cOmma', output is comma-separated. Basically useless.
=item 'l' : Long form of verb name is used.
=item 'x' : 'regeXp', when multiple forms are available, returns a
regexp that matches all correct form, and only these. This is mostly
useful for past participle of abundant verbs.
=item 'i' : 'ascIi', use only ascii characters, no accents.
=item 'h' : 'Hash', return a hash rather than a string. The keys are
the tense names (both long and short). The values are refs to arrays
whose i'th element is the i'th person, for i in 1-6. For example,
C<$h->{pres}->[1]> is the first person singular. C<$h->{pres}->[0]> is
not defined. For the particípio passado, only $h->{pp}->[1] is
defined, and likewise for gerundivo. This options renders ineffective
all other options except 'i' and 'x'.
=back
=head2 VERB
The verb argument must end in 'ar', 'or' '\^or', 'er' or 'ir'
(otherwise it cannot be a genuine portuguese verb). If a string with a
valid ending is passed, that is not a true portuguese verb -say,
'PerlProgrammer' - it will be conjugated as if it were a regular verb.
=head2 TENSE
The optional 'tense' arguments consist in a list of
'pres' , 'Presente',
'perf' , 'Perfeito',
'imp' , 'Imperfeito',
'fut' , 'Futuro',
'mdp' , 'Mais Que Perfeito' or 'Mais-que-Perfeito',
'cpres', 'Conjuntivo Presente',
'cimp' , 'Conjuntivo Imperfeito',
'cfut' , 'Conjuntivo Futuro',
'ivo' , 'Imperativo',
'pp' , 'Particípio Passado',
'grd' , 'Gerundivo'.
Caps don't matter. Default is all tenses in the first column.
=head2 PERSON
The optional 'person' arguments consist in a list of numbers
representing the person at which the verb is conjugated : 'I'=1,
'you'=2, 'she/he'=3, 'we'=4, 'you(plural)'=5 'they'=6. All numbers
other that 1-6 are ignored. Default is 1-6.
Now, for the inner details :
=head1 DATA STRUCTURES :
=head2 Variable C<$Lingua::PT::Conjugate::vlist>
The database is a single big string, defined at the end of
Conjugate.pm (the comments and end-of-line's are taken out).
It is in the variable C<$Lingua::PT::Conjugate::vlist>.
To understand the format, have a look at the code, and read this
informal description :
I<Definition of the conjugation of a verb>
<verb_name> ":" <tense>? <conjugated_form> <conjugated_form>...
The <tense> is optional : By default, the "parser" expects to find
them in the order pres (i.e. present), perf (perfeito), imp
(imperfeito), fut (futuro), mdp (mais-que-perfeito), cpres (conjuntivo
presente), cimp (conjuntivo imperfeito), cfut (conjuntivo futuro), cond
(condicional), ivo (imperativo), pp (particípio passado) and grd
(gerundivo).
The conjugated forms must correspond to 1st person, second, third,
first plural, second plural and third plural; except for
the imperative tense (shortname : "ivo") which starts at 2nd; and past
participle ("pp") and gerundivo ("grd") that have a single form
only. The conjugated forms may be regexes that match a correct form
(and only a correct form), when more than one form is ok (this is the
case of the "verbos abundivos").
The second person plural has been added in version 1.02, and is
probably buggy.
Shorthand notation may be used :
An "etc" means that the rest of that tense is regular.
A "." means that the current person is regular.
A "x" means that this verb is defective, and that the current
person is missing.
A "acc" means that circumflex and acute accents are toggled for the
current tense.
I<Saying what verbs follows the model of a given verb>
<model_name> = <verb_name> <verb_name> ...
Mean that all the verbs after the "=" have the verb on the lhs as
model.
I<Saying what verbs is defective>
defectivos1= <verb_name> ...
defectivos2= <verb_name> ...
defectivos3= <verb_name> ...
entries specifies defective verbs for which the function L<conjug>
has a hard-wired behavior.
=head2 C<%verb> is a hash in the format :
$verb->{$verb_name}->{$tense_name}->[$person] == "whole word".
This is used for irregular verbs. All the
$verb->{$verb_name}->{$tense_name}->[$person] and
$verb->{$verb_name}->{$tense_name} need not be defined.
If a verb follows a model (i.e. is conjugated like a verb which is
present in %verb's keys, and more completely defined), it has a
"model" key :
$verb->{$verb_name}->{"model"} == "model name".
If a verb is "defectivo" (has missing forms), but otherwise regular,
it will have an entry
$verb->{"defectivo"}->{$verb_name} == number.
where the number (1,2, or 3) defines how it behaves.
=head2 C<%reg> is a hash describing the ending of regular verbs :
$reg->{$verb_ending}->{$tense_name}->[$person] == "ending".
The verb ending may be [aeio]r.
=head2 C<%endg> contains regexes that match
verb endings, and is, right now, a little obsolete in the code.
=head2 C<codify($string)> reads a string in the
L<$vlist> format, and modifies the internal L<%verb> variable so that
it reflect the information if C<$string>.
=head2 C<&locate($verbname)> If $verbname is not a key of "%verb",
finds the data relevant to that verb in the string L<$vlist> , and
codes it so that L<%verb> reflects that data.
=head1 ALGORITHM of the C<conjug> sub :
Pick up options and arguments.
FOREACH verb $v,
Find the $root and ending ($edg).
Check if $v will require a special treatment, e.g. if it ends in
/g[ei]r$/, /c[ei]r$/, etc... and if it is the case, define a
"modifier" function (called &$modif), that will be called to
"final-polish" the output.
IF $v is IRREGULAR, e.g. if %verb{$v} is defined.
IF $v follows a model, find that model, and "see what it looks
like" (e.g. does the model's root ($rm) end with consonnants,
etc).
END
FOREACH tense $t and person $p ($b is in 1-4,6),
IF this form (e.g. $verb->{$v}->{$t}->[$p-1]) is explicitely
defined,
Great! That's it.
ELSE IF the model is explicitly defined, for this form,
"See how the model relates to the form, and how the verb $v
relates to the model, and from that find what should be $v's
corresponding form" (quite heuristic).
ELSE
Find the appropriate form for regular-irregular verbs (another
series of heuristics).
END
Check for "defective" verbs.
END
ELSE (verb is REGULAR)
It's trivial.
END
END That's it!
The actual code is much, much hairier.
=head2 WHAT MAKES ME BELIEVE IT WORKS?
I checked in the book "Guia Prático dos Verbos Portugueses", by Deolinda
Monteiro e Beatriz Pessoa. Well, to some extent ...
The file C<reference> reference contains conjugation tables for
plenty of verbs. Each time I modify C<Conjugate.pm>, I run the program
C<ckconj> ("ckconj all") that checks that the output is still correct
for all these verbs.
This program outputs two lines, one starting with "OK", containing
the verbs that checked ok, and a line "NOT IN REFERENCE FILE"
containing the verbs that "Conjugate.pm" knows about, but that are not
in the "reference" file.
Any other output reports found errors.
Since C<reference> is produced by C<conjug>, it may be inadvertely
contaminated. When I discover that, I hack C<Conjugate.pm> until it
conjugates correctly the faulty verb (by checking myself), *and all
the other ones too* (with "ckconj all")(like that, I check that my fix
has not broken anything). And then I run the program F<assert> (syntax
is "assert faulty_verb"), which modifies the reference table.
=head1 SEE ALSO : treinar, conjug.
=head1 VERSION 1.0
=head1 AUTHOR Etienne Grossmann, 1998-2006 [etienne@isr.ist.utl.pt]
=head1 CREDITS
Thanks to all people on usenet and here at ISR with whom I have
discussed about this module, who have provided advice on conjugation,
programming, on naming and on all relevant points.
Thanks to Lupe Christoph <lupe@lupe-christoph.de> from
cpan-testers@perl.org and to Miguel Marques
<marques@physik.uni-wuerzburg.de> for finding and fixing some bugs.
=cut