blob: ab4cabe4ed36d062a9720d60fc7c73c369fa10d8 [file] [log] [blame]
package Kalamar::API;
use Mojo::Base 'Mojolicious::Plugin';
use Scalar::Util qw/blessed weaken/;
use strict;
use warnings;
# KorAP Search engine for Mojolicious::Plugin::Search
# TODO: Add fixtures
# TODO: Support search in corpus and virtualcollection
# TODO: Support caching everywhere!
# TODO: Correct use of stash info everywhere!
# TODO: Alot is now underneath "meta"
# Register the plugin
sub register {
my ($plugin, $mojo, $index_class, $param) = @_;
$param ||= {};
# Add attributes to the index class
$index_class->attr(api => $param->{api});
$index_class->attr([qw/cutoff
query_language
time_exceeded
api_request
_api_cache
api_response
benchmark
query_jsonld/]);
$index_class->attr(no_cache => 0);
};
# Search the index
sub search {
my $self = shift;
my $index = shift;
# Get controller
my $c = $index->controller;
# If there is a callback, do async
my $cb = pop if ref $_[-1] && ref $_[-1] eq 'CODE';
# No query defined
unless ($index->query) {
return $cb->($index) if $cb;
return;
};
# Get query url
my $url = _query_url($index, @_);
# Cache based on URL
$index->_api_cache('total-' . $url->to_string);
my %param = @_;
# Set context based on parameter
$url->query({ context => $param{'context'} // 'paragraph' });
# Set path to search
$url->path('search');
# Check cache for total results
my $total_results;
if (!$index->no_cache &&
defined ($total_results = $c->chi->get($index->_api_cache))) {
# Set total results from cache
$index->total_results($total_results);
$c->app->log->debug('Get total result from cache');
# Set cutoff unless already set
$url->query({cutoff => 'true'}) unless defined $index->cutoff;
};
# Set api request for debugging
$index->api_request($url->to_string);
# Create new user agent and set timeout to 2 minutes
my $ua = $c->ua;
$ua->inactivity_timeout(120);
# Debugging
$c->app->log->debug('Search for ' . $index->api_request);
# Search non-blocking
if ($cb) {
$ua->get(
$url => sub {
my $tx = pop;
$self->_process_response('matches', $index, $tx);
weaken $index;
return $cb->($index);
});
}
# Search blocking
else {
my $tx = $ua->get($url);
$self->_process_response('matches', $index, $tx);
return $index;
};
};
# Trace query serialization
sub trace {
my $self = shift;
my $index = shift;
# Get controller
my $c = $index->controller;
# If there is a callback, do async
my $cb = pop if ref $_[-1] && ref $_[-1] eq 'CODE';
my %param = @_;
# No query defined
unless ($index->query(delete $param{query})) {
return $cb->($index) if $cb;
return;
};
# Get query url
my $url = _query_url($index, @_);
$url->path('search');
# Create new user agent and set timeout to 30 seconds
my $ua = $c->ua; # Mojo::UserAgent->new;
$ua->inactivity_timeout(30);
# Build transaction
my $tx = $ua->build_tx(TRACE => $url);
# non-blocking
if ($cb) {
weaken $index;
# Trace non-blocking
$ua->start(
$tx => sub {
$self->_process_response('trace', $index, pop);
return $cb->($index);
});
}
# Trace blocking
else {
my $tx = $ua->start($url);
return $self->_process_response('trace', $index, $tx);
};
};
# Get match info
sub match {
my $self = shift;
my $index = shift;
# Get controller
my $c = $index->controller;
# If there is a callback, do async
my $cb = pop if ref $_[-1] && ref $_[-1] eq 'CODE';
my %param = @_;
my $url = Mojo::URL->new($index->api);
# Legacy: In old versions, doc_id contained text_id
$param{doc_id} .= '.' . $param{text_id} if $param{text_id};
# Use hash slice to create path
$url->path(join('/', 'corpus', @param{qw/corpus_id doc_id match_id/}, 'matchInfo'));
# Build match id
# $match = 'match-' . $corpus . '!' . $corpus . '_' . $doc . '-' . $match;
my %query;
$query{foundry} = $param{foundry};
$query{layer} = $param{layer} if defined $param{layer};
$query{spans} = $param{spans} ? 'true' : 'false';
# Add query
$url->query(\%query);
$c->app->log->debug('Match info: ' . $url);
# Create new user agent and set timeout to 30 seconds
my $ua = $c->ua; # Mojo::UserAgent->new;
$ua->inactivity_timeout(30);
# non-blocking
if ($cb) {
weaken $index;
$ua->get(
$url => sub {
my $tx = pop;
$self->_process_response('match', $index, $tx);
return $cb->($index);
});
}
# Match info blocking
else {
my $tx = $ua->get($url);
return $self->_process_response('match', $index, $tx);
};
};
# Get resource information
sub resource {
my $self = shift;
my $index = shift;
# Get controller
my $c = $index->controller;
# If there is a callback, do async
my $cb = pop if ref $_[-1] && ref $_[-1] eq 'CODE';
my %param = @_;
# Rename info endpoints regarding resource
my $type = $param{type} // 'collection';
$type = 'virtualcollection' if $type eq 'collection';
# Create resource URL
my $url = Mojo::URL->new($index->api)->path($type);
# Debugging
$c->app->log->debug('Get resource info on '. $url);
# Check for cached information
if (my $json = $c->chi->get($url->to_string)) {
# TODO: That's unfortunate, as it prohibits caching of multiple resources
$c->app->log->debug('Get resource info from cache');
$c->stash('search.resource' => $json);
return $cb->($index) if $cb;
return $json;
};
$c->stash('search._resource_cache' => $url->to_string);
# Create new user agent and set timeout to 30 seconds
my $ua = $c->ua; # Mojo::UserAgent->new;
$ua->inactivity_timeout(30);
# Get resource information async
if ($cb) {
weaken $index;
$ua->get(
$url => sub {
$self->_process_response('resource', $index, pop);
return $cb->($index);
})
}
# Get resource information blocking
else {
my $tx = $ua->get($url);
$self->_process_response('resource', $index, $tx);
};
};
# Process response - especially error messages etc.
sub _process_response {
my ($self, $type, $index, $tx) = @_;
my $c = $index->controller;
# An error has occurded
if (my $e = $tx->error) {
$c->notify(
error =>
($e->{code} ? $e->{code} . ': ' : '') .
$e->{message} . ' for ' . $type . ' (remote)'
);
return;
};
# Response was fine
if (my $res = $tx->success) {
# Set api response for debugging
$index->api_response($res->body); # if $c->kalamar_test_port;
# Json failure
my $json;
unless ($json = $res->json) {
$c->notify(error => 'JSON response is invalid');
return;
};
# expected response for matches
if ($type eq 'matches') {
$self->_process_response_matches($index, $json);
}
elsif ($type eq 'trace') {
$self->_process_response_trace($index, $json);
}
elsif ($type eq 'match') {
$self->_process_response_match($index, $json);
}
elsif ($type eq 'resource') {
$self->_process_response_resource($index, $json);
};
return 1 if ref $json ne 'HASH';
$self->_notify_on_warnings($c, $json);
$self->_notify_on_error($c, 0, $json);
}
# Request failed
else {
$self->_notify_on_error($c, 1, $tx->res);
};
return 1;
};
# Handle match results
sub _process_response_matches {
my ($self, $index, $json) = @_;
# Reformat benchmark counter
my $benchmark = $json->{benchmark};
if ($benchmark && $benchmark =~ s/\s+(m)?s$//) {
$benchmark = sprintf("%.2f", $benchmark) . ($1 ? $1 : '') . 's';
};
# Set benchmark
$index->benchmark($benchmark);
# Set time exceeded
if ($json->{timeExceeded} && $json->{timeExceeded} eq Mojo::JSON::true) {
$index->time_exceeded(1);
};
# Set result values
$index->items_per_page($json->{itemsPerPage});
$index->query_jsonld($json->{request}->{query});
$index->results(_map_matches($json->{matches}));
# Total results not set by stash
if ($index->total_results == -1) {
if ($json->{totalResults} && $json->{totalResults} > -1) {
my $c = $index->controller;
$c->app->log->debug('Cache total result');
$c->chi->set($index->_api_cache => $json->{totalResults}, '120min');
$index->total_results($json->{totalResults});
};
};
};
# Process query serialization response
sub _process_response_match {
my ($self, $index, $json) = @_;
$index->results(_map_match($json));
};
# Process trace response
sub _process_response_trace {
my ($self, $index, $json) = @_;
$index->query_jsonld($json);
};
# Process resource response
sub _process_response_resource {
my ($self, $index, $json) = @_;
my $c = $index->controller;
# TODO: That's unfortunate, as it prohibits multiple resources
$c->stash('search.resource' => $json);
$c->app->log->debug('Cache resource info');
$c->chi->set($c->stash('search._resource_cache') => $json, '24 hours');
};
# Parse error messages and forward them to the user
sub _notify_on_error {
my ($self, $c, $failure, $res) = @_;
my $json = $res;
my $log = $c->app->log;
# Check if the response is already json
if (blessed $res) {
$json = $res->json if blessed $res ne 'Mojo::JSON';
};
# Check json response error message
if ($json) {
if ($json->{error}) {
# Temp
$json->{error} =~ s/;\s+null$//;
$c->notify(error => $json->{error});
return;
}
# New error messages
elsif ($json->{errstr}) {
# Temp
$json->{errstr} =~ s/;\s+null$//;
$c->notify(error => $json->{errstr});
return;
}
elsif ($json->{errors}) {
my $errors = $json->{errors};
# TODO: Check for ref!
foreach (@$errors) {
$c->notify(
error =>
($_->[0] ? $_->[0] . ': ' : '') .
$_->[1]
);
};
}
# policy service error messages
elsif ($json->{status}) {
$c->notify(error => 'Middleware error ' . $json->{status});
return;
};
};
# Doesn't matter what - there is a failure!
if ($failure) {
$c->notify(error => (
($res->{code} ? $res->{code} . ': ' : '') .
($res->{message} ? $res->{message} : 'Unknown error') .
' (remote)'
));
};
};
sub _notify_on_warnings {
my ($self, $c, $json) = @_;
# Add warnings (Legacy)
if ($json->{warning}) {
$json->{warning} =~ s/;\s+null$//;
$c->notify(warn => $json->{warning});
}
# Add warnings
elsif ($json->{warnings}) {
my $warnings = $json->{warnings};
# TODO: Check for ref!
foreach (@$warnings) {
$c->notify(
warn =>
($_->[0] ? $_->[0] . ': ' : '') .
$_->[1]
);
};
};
};
# Cleanup array of matches
sub _map_matches {
return () unless $_[0];
map { _map_match($_) } @{ shift() };
};
# Cleanup single match
sub _map_match {
my $x = shift or return;
$x->{ID} =~ s/^match\-[^!]+![^-]+-//;
$x->{docID} =~ s/^[^_]+_//;
# Legacy: In old versions the text_id was part of the doc_id
unless ($x->{textID}) {
($x->{docID}, $x->{textID}) = split '\.', $x->{docID};
};
$x;
};
# Build query url
sub _query_url {
my ($index, %param) = @_;
# Set cutoff from param
$index->cutoff(delete $param{cutoff});
# Set query language
$index->query_language(delete $param{query_language} // 'poliqarp');
# Should results be cached? Defaults to "yes"
$index->no_cache(1) if $param{no_cache};
# Init the query with stuff coming from the index
my %query;
$query{q} = $index->query;
$query{ql} = $index->query_language;
$query{page} = $index->start_page if $index->start_page;
$query{count} = $index->items_per_page if $index->items_per_page;
$query{cutoff} = 'true' if $index->cutoff;
# Todo: support corpus and collection
# Create query url
my $url = Mojo::URL->new($index->api);
$url->query(\%query);
return $url;
};
1;
__END__
=pod
=encoding utf8
=head1 NAME
Kalamar::API
=head1 DESCRIPTION
L<Kalamar::API> is a search engine class for L<Mojolicious::Plugin::Search>
that uses the KorAP Web API.
B<The Web API as well as L<Mojolicious::Plugin::Search> are not stable yet,
so this class is expected to change in the near future. Do not rely on its API!>
=head1 METHODS
L<Kalamar::API> inherits all methods from L<Mojolicious::Plugin> and
implements the following new ones.
=head2 register
See L<Mojolicious::Plugin::Search> for registering search engines.
In addition to the mentioned query parameters, the following parameters are supported:
=over 2
=item B<query_language>
One of the supported query languages, like C<poliqarp> or C<annis>.
=item B<cutoff>
Cut off results following the current page (i.e. don't count the number of results).
=item B<no_cache>
Do not cache search results. Defaults to C<0>.
=back
In addition to the mentioned index attributes, the following attributes are supported:
=over 2
=item B<api>
The API address.
=item B<time_exceeded>
Report on time outs, that may mean, not all results were retrieved.
=item B<api_request>
Report the whole API request.
=item B<api_response>
Report the whole API response (a KoralQuery object).
=item B<benchmarks>
Report on processing time for benchmarking.
=item B<query_jsonld>
The KoralQuery realization of the C<query> object.
=back
=head2 search
Search the index.
=head2 trace
Trace query serializations.
=head2 match
Get match information.
=head2 resource
Get resource information.
=head1 COPYRIGHT AND LICENSE
Copyright (C) 2015, L<IDS Mannheim|http://www.ids-mannheim.de/>
Author: L<Nils Diewald|http://nils-diewald.de/>
Kalamar is developed as part of the L<KorAP|http://korap.ids-mannheim.de/>
Corpus Analysis Platform at the
L<Institute for the German Language (IDS)|http://ids-mannheim.de/>,
member of the
L<Leibniz-Gemeinschaft|http://www.leibniz-gemeinschaft.de/en/about-us/leibniz-competition/projekte-2011/2011-funding-line-2/>
and supported by the L<KobRA|http://www.kobra.tu-dortmund.de> project,
funded by the
L<Federal Ministry of Education and Research (BMBF)|http://www.bmbf.de/en/>.
Kalamar is free software published under the
L<BSD-2 License|https://raw.githubusercontent.com/KorAP/Kalamar/master/LICENSE>.
=cut