Algorithm-Classifier-IsolationForest
view release on metacpan or search on metacpan
lib/Algorithm/Classifier/IsolationForest/App/Command/streamd.pm view on Meta::CPAN
't=s@',
'Feature name tag. Pass once per feature; enables the tagged (JSON object) row form '
. '(new models only).'
],
[
'mungers=s',
'JSON file of Algorithm::ToNumberMunger specs, keyed by feature tag (new models only; requires -t).',
{ 'completion' => 'files' }
],
[
'prototype=s',
'JSON prototype file to create the model from (new models only). May not be combined '
. 'with -t or --mungers. See PROTOTYPES in the module POD.',
{ 'completion' => 'files' }
],
);
} ## end sub opt_spec
sub abstract { 'Run an Online Isolation Forest scoring daemon on a Unix socket, speaking JSON lines' }
sub description {
'Runs a prequential scoring daemon around an Online Isolation Forest
model (Algorithm::Classifier::IsolationForest::Online): clients connect
to the Unix domain socket and exchange one JSON document per line.
At startup the daemon resumes from <model-dir>/latest.json when it
exists; otherwise it creates a new model from the creation knobs (-n,
--window, --eta, --growth, --subsample, -s, -c, -t, --mungers,
--prototype -- the same set `iforest stream` takes). The model is saved
to a timestamped file in --model-dir every --save-interval seconds
(only when something was learned), on SIGUSR1, on the save command, and
at shutdown; the symlink latest.json is atomically repointed at every
save, so a restart resumes the stream losing at most one interval.
Requests are JSON objects carrying exactly one of "row", "rows", or
"cmd", an optional "mode", and an optional "tag" (any JSON value,
echoed back verbatim in the reply -- a correlation tag, not to be
confused with feature tags):
{"row": [0.1, 0.7]} -> {"score": 0.41, "label": 0}
{"row": {"cpu": 0.1, "mem": 0.7}} -> {"score": 0.41, "label": 0}
{"rows": [[...], {...}], "tag": "b7"} -> {"scores": [[0.41,0], ...], "tag": "b7"}
{"rows": [[...]], "mode": "learn"} -> {"ok": {"learned": 1}}
{"cmd": "mode", "mode": "score"} -> {"ok": {"mode": "score"}}
{"cmd": "ping"} -> {"ok": "pong"}
{"cmd": "stats"} -> {"ok": {"seen": ..., ...}}
{"cmd": "save"} -> {"ok": {"saved": "oiforest-....json"}}
{"cmd": "relearn-threshold"} -> {"ok": {"threshold": 0.61}}
anything invalid -> {"error": "...", "tag": ...}
The array row form is positional (scalar mungers applied, like stream
CSV input); the object form is a tagged row and runs the full munger
plan, including expanding and combining mungers -- and, being JSON, the
raw values may safely contain commas, newlines, or any unicode.
A worked tagged example. Create the daemon around raw HTTP request
data, with mungers turning the raw values into numbers (mungers.json
here; a --prototype carrying the same schema works identically):
{ "method": { "munger": "http_method_enum", "default": -1 },
"path_len": { "munger": "length", "from": "path" },
"host_entropy": { "munger": "entropy", "from": "host" } }
iforest streamd --set web -t method -t path_len -t host_entropy \
--mungers mungers.json -c 0.05
Clients then send the raw values themselves -- note the input fields
are the munger SOURCES (method, path, host), not the feature tags,
because the plan derives path_len and host_entropy from them:
-> {"row": {"method": "GET", "path": "/index.html",
"host": "www.example.com"}, "tag": "r-1"}
<- {"score": 0.31, "label": 0, "tag": "r-1"}
-> {"row": {"method": "BREW", "path": "/aa,a\"a.php",
"host": "kq3xv9z2.biz"}, "tag": "r-2"}
<- {"score": 0.74, "label": 1, "tag": "r-2"}
The same rows work from the shell via
`iforest streamc --set web --jsonl -i rows.jsonl`.
Modes are prequential (score each row against the model as it stood, then
learn it -- the default), learn (learn only), and score (score only);
"mode" on a row/rows message overrides the connection default set by
the mode command for that message. A bad row gets an {"error": ...}
reply on that message only; the connection and the daemon live on (for
a "rows" batch, rows before the failing one were already processed).
Multiple concurrent connections are supported; rows are applied to the
one shared model in the order their lines arrive, which defines the
stream order.
--set NAME runs a named instance: the set name is appended to
--model-dir (so its saves, latest.json, and default log live under
their own subdirectory) and the socket/pid become <set>.sock /
<set>.pid under the run dir -- with --set, --socket and --pid name the
base run dir instead of the files. Several sets run side by side, each
with its own model, resume state, and double-start protection:
iforest streamd --set web
iforest streamd --set dns --prototype dns-proto.json -c 0.02
Set names must match /\A[A-Za-z0-9+\-@_]+\z/; since the class has no
"." or "/", a set name can only ever create one new path segment.
Everything under --model-dir and the socket/pid directories is created
at startup when missing; when that fails (e.g. running unprivileged
with the /var defaults) the daemon dies immediately, before forking,
naming the directory and the flag to override.
';
} ## end sub description
sub validate {
my ( $self, $opt, $args ) = @_;
# Anchored with \A/\z rather than ^/$ ($ tolerates a trailing newline).
# The class has no '.' or '/', so a set name can only ever create one
# new path segment -- no traversal is expressible.
if ( defined( $opt->{'set'} ) && $opt->{'set'} !~ /\A[A-Za-z0-9+\-@_]+\z/ ) {
$self->usage_error( '--set, "'
. $opt->{'set'}
. '", must match /\A[A-Za-z0-9+\-@_]+\z/ (letters, digits, and + - @ _ only)' );
lib/Algorithm/Classifier/IsolationForest/App/Command/streamd.pm view on Meta::CPAN
} elsif ( !defined( $opt->{'t'} ) ) {
$self->usage_error('--mungers requires feature tags (-t) to compile against');
}
}
if ( defined( $opt->{'prototype'} ) ) {
if ( !-f $opt->{'prototype'} ) {
$self->usage_error( '--prototype, "' . $opt->{'prototype'} . '", is not a file or does not exist' );
} elsif ( !-r $opt->{'prototype'} ) {
$self->usage_error( '--prototype, "' . $opt->{'prototype'} . '", is not readable' );
}
if ( defined( $opt->{'t'} ) || defined( $opt->{'mungers'} ) ) {
$self->usage_error(
'--prototype may not be combined with -t or --mungers; the schema comes only from the prototype');
}
} ## end if ( defined( $opt->{'prototype'} ) )
return 1;
} ## end sub validate
sub execute {
my ( $self, $opt, $args ) = @_;
# JSON::MaybeXS is required lazily so a box without it still has a
# working iforest CLI (App::Cmd loads every command module up front).
eval { require JSON::MaybeXS; 1 }
or die( 'iforest streamd requires JSON::MaybeXS for its wire protocol; install it: ' . $@ );
$JSON = JSON::MaybeXS->new( utf8 => 1, canonical => 1, allow_nonref => 0 );
%OPT = %$opt;
# --set turns --socket/--pid into base run dirs holding <set>.sock /
# <set>.pid and appends the set name to --model-dir, so several named
# daemons run side by side with no other flags. Without a set the
# flags are the socket/pid files themselves, defaulting as documented.
if ( defined $OPT{'set'} ) {
my $run = defined $OPT{'socket'} ? $OPT{'socket'} : '/var/run/iforest_streamd';
$OPT{'socket'} = File::Spec->catfile( $run, $OPT{'set'} . '.sock' );
my $prun = defined $OPT{'pid'} ? $OPT{'pid'} : '/var/run/iforest_streamd';
$OPT{'pid'} = File::Spec->catfile( $prun, $OPT{'set'} . '.pid' );
$OPT{'model_dir'} = File::Spec->catdir( $OPT{'model_dir'}, $OPT{'set'} );
} else {
$OPT{'socket'} = '/var/run/iforest_streamd/streamd.sock' unless defined $OPT{'socket'};
$OPT{'pid'} = '/var/run/iforest_streamd/streamd.pid' unless defined $OPT{'pid'};
}
# Daemonizing chdirs to /, so every path used after that point must be
# absolute -- including the socket and pid file, which are unlinked at
# shutdown.
for my $path_opt (qw(socket pid model_dir log)) {
$OPT{$path_opt} = File::Spec->rel2abs( $OPT{$path_opt} )
if defined $OPT{$path_opt};
}
# sun_path is 104 bytes on the BSDs and 108 on Linux (including the
# NUL); Socket.pm just warns and TRUNCATES an over-long path, which
# binds a socket nobody will ever find. Refuse loudly instead.
die( '--socket, "'
. $OPT{'socket'}
. '", is '
. length( $OPT{'socket'} )
. ' bytes; Unix socket paths are limited to ~104 bytes -- use a shorter path'
. "\n" )
if length( $OPT{'socket'} ) > 100;
# --- directories, before anything forks or binds -----------------------
_ensure_dir( $OPT{'model_dir'}, '--model-dir' );
_ensure_dir( dirname( $OPT{'socket'} ), '--socket' );
_ensure_dir( dirname( $OPT{'pid'} ), '--pid' );
# --- refuse to double-start ---------------------------------------------
if ( -e $OPT{'socket'} ) {
my $probe = IO::Socket::UNIX->new( Peer => $OPT{'socket'} );
die( 'another daemon is already listening on "' . $OPT{'socket'} . '"' . "\n" ) if $probe;
unlink $OPT{'socket'}; # stale socket from an unclean exit
}
if ( -f $OPT{'pid'} ) {
my $old = read_file( $OPT{'pid'} );
chomp $old if defined $old;
if ( defined $old && $old =~ /\A\d+\z/ && ( kill( 0, $old ) || $!{EPERM} ) ) {
die( 'another daemon appears to be running (pid ' . $old . ' from "' . $OPT{'pid'} . '")' . "\n" );
}
unlink $OPT{'pid'}; # stale pid file
}
# --- resume or create the model ----------------------------------------
my $latest = File::Spec->catfile( $OPT{'model_dir'}, 'latest.json' );
if ( -e $latest ) {
$OIF = Algorithm::Classifier::IsolationForest->load($latest);
die( '"' . $latest . '" is not an online model; streamd only works on those' . "\n" )
unless ref $OIF eq 'Algorithm::Classifier::IsolationForest::Online';
} elsif ( defined $OPT{'prototype'} ) {
my $proto = eval {
Algorithm::Classifier::IsolationForest->validate_prototype( scalar read_file( $OPT{'prototype'} ) );
};
die( '--prototype, "' . $OPT{'prototype'} . '", is not a valid prototype: ' . $@ ) if $@;
die( '--prototype, "' . $OPT{'prototype'} . '", is for a batch model; streamd needs an online one' . "\n" )
unless $proto->{class} eq 'online';
my %overrides;
$overrides{'n_trees'} = $OPT{'n'} if defined $OPT{'n'};
$overrides{'window_size'} = $OPT{'window'} if defined $OPT{'window'};
$overrides{'max_leaf_samples'} = $OPT{'eta'} if defined $OPT{'eta'};
$overrides{'growth'} = $OPT{'growth'} if defined $OPT{'growth'};
$overrides{'subsample'} = $OPT{'subsample'} if defined $OPT{'subsample'};
$overrides{'seed'} = $OPT{'s'} if defined $OPT{'s'};
$overrides{'contamination'} = $OPT{'c'} if defined $OPT{'c'};
$OIF = eval { Algorithm::Classifier::IsolationForest->new_from_prototype( $proto, %overrides ) };
die( '--prototype, "' . $OPT{'prototype'} . '", failed to create a model: ' . $@ ) if $@;
} else {
my $mungers;
if ( defined $OPT{'mungers'} ) {
$mungers = eval { $JSON->decode( scalar read_file( $OPT{'mungers'} ) ) };
die( '--mungers, "' . $OPT{'mungers'} . '", did not parse as JSON: ' . $@ ) if $@;
die( '--mungers, "' . $OPT{'mungers'} . '", must be a JSON object of tag => spec' )
unless ref $mungers eq 'HASH';
}
$OIF = Algorithm::Classifier::IsolationForest::Online->new(
'n_trees' => $OPT{'n'},
'window_size' => $OPT{'window'},
'max_leaf_samples' => $OPT{'eta'},
'growth' => $OPT{'growth'},
'subsample' => $OPT{'subsample'},
lib/Algorithm/Classifier/IsolationForest/App/Command/streamd.pm view on Meta::CPAN
my $base = 'oiforest-' . strftime( '%Y%m%d-%H%M%S', localtime );
my $name = $base . '.json';
my $n = 0;
while ( -e File::Spec->catfile( $OPT{'model_dir'}, $name ) ) {
$n++;
$name = $base . '-' . $n . '.json';
}
write_file( File::Spec->catfile( $OPT{'model_dir'}, $name ), { 'atomic' => 1 }, $OIF->to_json );
my $tmp = File::Spec->catfile( $OPT{'model_dir'}, '.latest.tmp.' . $$ );
unlink $tmp;
symlink( $name, $tmp )
or _log( 'WARNING: symlink for latest.json failed: ' . $! );
rename( $tmp, File::Spec->catfile( $OPT{'model_dir'}, 'latest.json' ) )
or _log( 'WARNING: renaming latest.json symlink failed: ' . $! );
$DIRTY = 0;
_log( 'saved ' . $name . ' (' . $why . ', seen=' . $OIF->seen . ')' );
_prune_models() if defined $OPT{'keep'};
return $name;
} ## end sub _save_model
sub _prune_models {
opendir( my $dh, $OPT{'model_dir'} ) or return;
my @models = sort { ( stat($a) )[9] <=> ( stat($b) )[9] }
map { File::Spec->catfile( $OPT{'model_dir'}, $_ ) }
grep { /\Aoiforest-.*\.json\z/ } readdir($dh);
closedir $dh;
while ( scalar @models > $OPT{'keep'} ) {
my $old = shift @models;
unlink $old and _log( 'pruned ' . $old );
}
return 1;
} ## end sub _prune_models
#-------------------------------------------------------------------------------
# connection handling
#-------------------------------------------------------------------------------
sub _drop {
my ( $s, $rsel, $wsel ) = @_;
$rsel->remove($s);
$wsel->remove($s);
delete $CONN{ fileno($s) };
close $s;
return 1;
}
sub _read_from {
my ( $s, $rsel, $wsel ) = @_;
my $c = $CONN{ fileno($s) } or return;
my $got = sysread( $s, my $chunk, 65536 );
if ( !defined $got ) {
return if $!{EAGAIN} || $!{EWOULDBLOCK} || $!{EINTR};
return _drop( $s, $rsel, $wsel );
}
return _drop( $s, $rsel, $wsel ) if $got == 0; # client closed
$c->{inbuf} .= $chunk;
if ( length( $c->{inbuf} ) > MAX_INBUF && $c->{inbuf} !~ /\n/ ) {
_log( 'dropping client: unterminated line exceeded ' . MAX_INBUF . ' bytes' );
return _drop( $s, $rsel, $wsel );
}
while ( $c->{inbuf} =~ s/\A([^\n]*)\n// ) {
my $line = $1;
next if $line =~ /\A\s*\z/;
_handle_line( $c, $line );
return _drop( $s, $rsel, $wsel ) if length( $c->{outbuf} ) > MAX_OUTBUF;
}
_flush( $s, $rsel, $wsel ) if length $c->{outbuf};
return 1;
} ## end sub _read_from
sub _flush {
my ( $s, $rsel, $wsel ) = @_;
my $c = $CONN{ fileno($s) } or return;
while ( length $c->{outbuf} ) {
my $wrote = syswrite( $s, $c->{outbuf} );
if ( !defined $wrote ) {
if ( $!{EAGAIN} || $!{EWOULDBLOCK} || $!{EINTR} ) {
$wsel->add($s) unless $wsel->exists($s);
return;
}
return _drop( $s, $rsel, $wsel );
}
substr( $c->{outbuf}, 0, $wrote, '' );
} ## end while ( length $c->{outbuf} )
$wsel->remove($s) if $wsel->exists($s);
return 1;
} ## end sub _flush
#-------------------------------------------------------------------------------
# protocol
#-------------------------------------------------------------------------------
# One request line -> one reply line, appended to the connection's
# output buffer. Any croak from the model becomes an {"error": ...}
# reply on this message alone; the connection and daemon live on.
sub _handle_line {
my ( $c, $line ) = @_;
my $msg = eval { $JSON->decode($line) };
if ( $@ || ref $msg ne 'HASH' ) {
( my $err = $@ ) =~ s/ at \S+ line \d+\.?\s*\z//s;
return _reply( $c, { error => 'request is not a JSON object: ' . ( $err || 'wrong type' ) } );
}
my @tag = exists $msg->{tag} ? ( tag => $msg->{tag} ) : ();
my @kinds = grep { exists $msg->{$_} } qw(row rows cmd);
if ( scalar @kinds != 1 ) {
return _reply( $c, { error => q{request needs exactly one of "row", "rows", or "cmd"}, @tag } );
}
my $kind = $kinds[0];
my $mode = exists $msg->{mode} ? $msg->{mode} : $c->{mode};
if ( !defined $mode || ref $mode || $mode !~ /\A(?:prequential|learn|score)\z/ ) {
return _reply( $c, { error => q{mode must be "prequential", "learn", or "score"}, @tag } );
}
if ( $kind eq 'cmd' ) {
return _handle_cmd( $c, $msg, \@tag );
}
my $rows = $kind eq 'row' ? [ $msg->{row} ] : $msg->{rows};
if ( ref $rows ne 'ARRAY' || !@$rows ) {
return _reply( $c, { error => q{"rows" must be a non-empty JSON array of rows}, @tag } );
}
my @scored;
my $i = 0;
for my $row (@$rows) {
my $score = eval { _apply_row( $row, $mode ) };
if ($@) {
( my $err = $@ ) =~ s/ at \S+ line \d+\.?\s*\z//s;
chomp $err;
my $where = $kind eq 'row' ? '' : 'row ' . $i . ': ';
return _reply( $c, { error => $where . $err, @tag } );
}
push @scored, $score if defined $score;
$i++;
} ## end for my $row (@$rows)
if ( $mode eq 'learn' ) {
return _reply( $c, { ok => { learned => scalar @$rows }, @tag } );
}
( run in 0.577 second using v1.01-cache-2.11-cpan-995e09ba956 )