pjs/tools/module-deps/module-graph.pl

529 строки
13 KiB
Perl
Executable File

#!/usr/bin/perl -w
# Usage:
# module-graph.pl [directory [ directory ..] ] > foo.dot
#
# Description:
# Outputs a Graphviz-compatible graph description file for use
# with the utilities dot, sccmap, and so forth.
# Graphviz is available from:
# http://www.research.att.com/sw/tools/graphviz/
#
# Reccomendations:
# View the graphs by creating graphs with dot:
# > dot -Tpng foo.dot -o foo.png
#
# Note to Linux users: graphviz needs TrueType fonts installed
# http://support.pa.msu.edu/Help/FAQs/Linux/truetype.html
#
# Todo:
# - eliminate arcs implied by transitive dependancies
# (i.e. in a -> b -> c; a->c;, eliminate a->c;
# (discovered that "tred" will do this, but isn't super-helpful)
# - group together strongly-connected components, where strongly connected
# means there exists a cycle, and all dependancies off the cycle.
# in the graph "a -> b <-> c -> d", b and c are strongly connected, and
# they depend on d, so b, c, and d should be grouped together.
use strict;
# For --option1, --option2, ...
use Getopt::Long;
Getopt::Long::Configure("bundling_override");
Getopt::Long::Configure("auto_abbrev");
sub PrintUsage {
die <<END_USAGE
Prints out required modules for specified directories.
usage: module-graph.pl [--start-module <mod> ] [--force-order <file> ] [--file <file> | <dir1> <dir2> ...] [--list-only] [--skip-tree] [--skip-dep-map] [--skip-list]
END_USAGE
}
my %clustered; # strongly connected components.
my %deps;
my %toplevel_modules; # visited components.
my %all_modules;
my $debug = 0;
my $makecommand = "make";
use Cwd;
my @dirs;
my $curdir = getcwd();
my $opt_list_only = 0; # --list-only: only print out module names
my $opt_dont_print_tree = 0; # --skip-tree: don't print dependency tree
my $opt_dont_print_dep_map = 0; # --skip-dep-map: don't print dependency map
my $opt_dont_print_list = 0; # --skip-list: don't print dependency list
my $load_file = 0; # --file
my $opt_start_module; # --start-module optionally print out dependencies
# for a given module.
my $force_order = 0; # --force-order gives rules on ordering outside of
# normal means.
# Parse commandline input.
sub parse_args() {
# Stuff arguments into variables.
# Print usage if we get an unknown argument.
PrintUsage() if !GetOptions('list-only' => \$opt_list_only,
'skip-dep-map' => \$opt_dont_print_dep_map,
'skip-tree' => \$opt_dont_print_tree,
'skip-list' => \$opt_dont_print_list,
'start-module=s' => \$opt_start_module,
'file=s' => \$load_file,
'force-order=s' => \$force_order);
# list-only: don't print tree + don't pring dep map
if($opt_list_only) {
$opt_dont_print_tree = 1;
$opt_dont_print_dep_map = 1;
}
# Last args are directories, if not in load-from-file mode.
unless($load_file) {
if (!@ARGV) {
@dirs = (getcwd());
} else {
@dirs = @ARGV;
# XXX does them in reverse order..
my $arg;
foreach $arg (@ARGV) {
push @dirs, "$curdir/$arg";
}
}
}
}
# Build up the %deps matrix.
sub build_deps_matrix() {
MFILE:
while ($#dirs != -1) {
my ($current_dirs, $current_module, $current_requires);
# pop the curdir
$curdir = pop @dirs;
if(!$opt_list_only) {
print STDERR "Entering $curdir.. \r";
}
chdir "$curdir" || next;
next if (! -e "$curdir/Makefile");
$current_dirs = "";
open(MAKEOUT, "$makecommand echo-dirs echo-module echo-requires|") || die "Can't make: $!\n";
$current_dirs = <MAKEOUT>; $current_dirs && chop $current_dirs;
$current_module = <MAKEOUT>; $current_module && chop $current_module;
$current_requires = <MAKEOUT>; $current_requires && chop $current_requires;
close MAKEOUT;
if ($current_module) {
#
# now keep a list of all dependencies of the module
#
my @require_list = split(/\s+/,$current_requires);
my $req;
foreach $req (@require_list) {
$deps{$current_module}{$req}++;
}
$toplevel_modules{$current_module}++;
$all_modules{$current_module}++;
}
next if !$current_dirs;
# now push all child directories onto the list
my @local_dirs = split(/\s+/,$current_dirs);
for (@local_dirs) {
push @dirs,"$curdir/$_" if $_;
}
}
if(!$opt_list_only) {
print STDERR "\n";
}
}
#
#
#
sub build_deps_matrix_from_file {
my ($filename) = @_;
open DEPS_FILE, $filename or print "can't open $filename, $?\n";
my @line;
while (<DEPS_FILE>) {
if(/->/) {
chomp;
s/^\s+//; # Strip off leading spaces.
s/\;//; # Strip off ';'
# Pick off module, and dependency from -> line.
@line = split(' -> ', $_);
$deps{$line[0]}{$line[1]}++;
# Add the module to the list of modules.
$toplevel_modules{$line[0]}++;
$all_modules{$line[0]}++;
} elsif (/style=filled/) {
chomp;
s/^\s+//; # Strip off leading spaces.
s/\;//; # Strip off ';'
# Pick off module
@line = split(' \[', $_);
$all_modules{$line[0]}++;
}
}
close DEPS_FILE;
}
# Print out %deps.
sub print_deps_matrix() {
my $module;
if(!$opt_dont_print_dep_map) {
print "digraph G {\n";
print " concentrate=true;\n";
# figure out the internal nodes, and place them in a cluster
#print " subgraph cluster0 {\n";
#print " color=blue;\n"; # blue outline around cluster
# ** new method: just list all modules that came from MODULE=foo
foreach $module (sort keys %toplevel_modules) {
print " $module [style=filled];\n"
}
}
print_dependency_list();
if(!$opt_dont_print_dep_map) {
print "}\n";
}
}
#
# Here we order an array based on other rules, and return the result.
# Optional from --force-order.
#
sub possibly_force_order {
my @modarray = @_;
if ($force_order) {
open FORCEORDER, $force_order or die "can't open force order file $force_order\n";
my @mod_orders;
LINE: while(<FORCEORDER>) {
# skip comments
next LINE if /^#/;
# Should be two strings, or we skip.
@mod_orders = split;
if( $#mod_orders + 1 == 2 ) {
my $mod_before = $mod_orders[0];
my $has_before = -1;
my $mod_after = $mod_orders[1];
my $has_after = -1;
my $traverse = 0;
my $mod_item = "";
foreach $mod_item (@modarray) {
# strip whitespace.
for ($mod_item) {
s/^\s+//;
s/\s+$//;
}
if( $mod_item eq $mod_before) {
$has_before = $traverse;
}
if( $mod_item eq $mod_after ) {
$has_after = $traverse;
}
$traverse++;
}
# Must find both strings.
# Must be out of order to rewrite string.
if(-1 != $has_before && -1 != $has_after && $has_before > $has_after) {
# Rewrite the array.
my $modules_string = "";
$traverse = 0;
foreach $mod_item (@modarray) {
# skip before, it comes with after.
if( $traverse != $has_before ) {
# if after, then we add before.
if( $traverse == $has_after ) {
$modules_string .= $mod_before;
$modules_string .= " ";
}
# add this item to the string.
$modules_string .= $mod_item;
$modules_string .= " ";
}
$traverse++;
}
# rewrite return value.
@modarray = split(' ', $modules_string);
}
}
}
close FORCEORDER;
}
return @modarray;
}
# ** old method: find only internal nodes
# (nodes with both parents and children)
sub print_internal_nodes() {
my $module;
my $depmod;
foreach $module (sort { scalar keys %{$deps{$b}} <=> scalar keys %{$deps{$a}} } keys %deps) {
foreach $depmod ( keys %deps ) {
# only in cluster if they are a child too
if ($deps{$depmod}{$module}) {
print " $module;\n";
$clustered{$module}++;
last;
}
}
}
}
# Run over dependency array to generate raw component list.
# This is the "a -> b" lines.
sub print_dependency_list() {
my @raw_list;
my @unique_list;
my $module;
foreach $module (sort sortby_deps keys %deps) {
my $req;
foreach $req ( sort { $deps{$module}{$b} <=> $deps{$module}{$a} }
keys %{ $deps{$module} } ) {
# print " $module -> $req [weight=$deps{$module}{$req}];\n";
if(!$opt_dont_print_dep_map) {
print " $module -> $req;\n";
} else {
# print "$req ";
push(@raw_list, $req);
}
}
}
# generate unique list, print it out.
if($opt_list_only) {
my %saw;
undef %saw;
@unique_list = grep(!$saw{$_}++, @raw_list);
# apply forced order if desired.
@unique_list = possibly_force_order(@unique_list);
my $i;
for ($i=0;$i <= $#unique_list; $i++) {
print $unique_list[$i], " ";
}
print "\n";
}
}
# we're sorting based on clustering
# order:
# - unclustered, with dependencies
# - clustered
# - unclustered, with no dependencies
# However, the last group will probably never come in $a or $b, because we're
# probably only being called from the keys in $deps
# We'll keep all the logic here, in case we come up with a better scheme later
sub sortby_deps() {
my $keys_a = scalar keys %{$deps{$a}};
my $keys_b = scalar keys %{$deps{$b}};
# determine if they are the same or not
if ($clustered{$a} && $clustered{$b}) {
# both in "clustered" group
return $keys_a <=> $keys_b;
}
elsif (!$clustered{$a} && !$clustered{$b}) {
# not clustered. Do they both have dependencies or both
# have no dependencies?
if (($keys_a && $keys_b) ||
(!$keys_a && !$keys_b)) {
# both unclustered, and either both have dependencies,
# or both don't have dependencies
return $keys_a <=> $keys_b;
}
}
# if we get here, then they are in different "groups"
if ($clustered{$a}) {
# b must be unclustered
if ($keys_b) {
return 1;
} else {
return -1;
}
} elsif ($clustered{$b}) {
# a must be unclustered
if ($keys_a) {
return -1;
} else {
return 1;
}
} else {
# both are unclustered, so the with-dependencies one comes first
if ($keys_a) {
return -1;
} else {
return 1;
}
}
}
#
# Recursively traverse the deps matrix.
#
my %visited_nodes;
my @visited_nodes_leaf_first_order; # Store in post-recursion order.
sub walk_module_digraph {
my ($module, $level) = @_;
# Remember that we visited this node.
$visited_nodes{$module}++;
# Print this node.
if (!$opt_dont_print_tree) {
my $i;
for ($i=0; $i<$level; $i++) {
print " ";
}
print "$module\n";
}
# If we haven't visited this node, search again
# from this node.
my $depmod;
foreach $depmod ( keys %{ $deps{$module} } ) {
my $visited = $visited_nodes{$depmod};
if(!$visited) { # test recursion: if($level < 5)
walk_module_digraph($depmod, $level + 1);
}
}
# Post-recursion. Store in array form so we keep the order.
push(@visited_nodes_leaf_first_order, $module);
if (!$opt_dont_print_tree) {
if($level == 1) {
print "\n";
}
}
}
sub print_module_deps {
# The "ALL" module requires special handling.
# Just print out the toplevel module names
if ($opt_start_module eq "ALL") {
my $key;
my @outlist;
foreach $key (keys %all_modules) {
push @outlist, $key;
}
if ($opt_list_only) {
print "@outlist\n";
return;
}
}
# Recursively hunt down dependencies for $opt_start_module
walk_module_digraph($opt_start_module, 1);
# apply forced ordering if needed.
@visited_nodes_leaf_first_order = possibly_force_order(@visited_nodes_leaf_first_order);
# Post-recursion version.
unless ($opt_dont_print_list) {
my $visited_mod;
foreach $visited_mod (@visited_nodes_leaf_first_order) {
print "$visited_mod ";
}
print "\n";
}
if($debug) {
my @total_visited = (sort keys %visited_nodes);
my $total = $#total_visited + 1;
print "\ntotal = $total\n";
}
}
sub get_matrix_size {
my (%matrix) = @_;
my $i;
my $j;
$i = 0;
$j = 0;
foreach $i ( keys %matrix ) {
$j++;
}
return $j;
}
# main
{
parse_args();
if($load_file) {
build_deps_matrix_from_file($load_file);
} else {
build_deps_matrix();
}
# Print out deps matrix.
# --list-only and --start-module together mean to
# print out the module deps, not the matrix.
if (not ($opt_list_only)) {
print_deps_matrix();
}
# If we specified a --start-module option, print out
# the required modules for that module.
if($opt_start_module) {
print_module_deps();
}
if($debug) {
print STDERR "----- sizes -----\n";
print STDERR " deps: " . get_matrix_size(%deps) . "\n";
print STDERR "toplevel_modules: " . get_matrix_size(%toplevel_modules) . "\n";
print STDERR " clustered: " . get_matrix_size(%clustered) . "\n";
}
}