On Tue, Jun 21, 2011 at 08:27:45PM -0400, Andrew Dunstan wrote:
>
> Attached is a WIP possible replacement for pgindent. Instead of a
> shell script invoking a mishmash of awk and sed, some of which is
> pretty impenetrable, it uses a single engine (perl) to do all the
> pre and post indent processing. Of course, if your regex-fu and
> perl-fu is not up the scratch this too might be impenetrable, but
> all but a couple of the recipes are reduced to single lines, and I'd
> argue that they are all at least as comprehensible as what they
> replace.
>
> Attached also is a diff file showing what it does differently from
> the existing script. I think that these are all things where the new
> script is more correct than the existing script. Most of the changes
> come into two categories:
>
> * places where the existing script fails to combine the function
> return type and the function name on a single line in function
> prototypes.
> * places where unwanted blank lines are removed by the new script
> but not by the existing script.
>
> Features include:
>
> * command line compatibility with the existing script, so you can do:
> find ../../.. -name '*.[ch]' -type f -print | egrep -v -f
> exclude_file_patterns | xargs -n100 ./pgindent.pl typedefs.list
> * a new way of doing the same thing much more nicely:
> ./pgindent.pl --search-base=../../.. --typedefs=typedefs.list
> --excludes=exclude_file_patterns
> * only passes relevant typedefs to indent, not the whole huge list
> * should in principle be runnable on Windows, unlike existing script
> (I haven't tested yet)
> * no semantic tab literals; tabs are only generated using \t and
> tested for using \t, \h or \s as appropriate. This makes debugging
> the script much less frustrating. If something looks like a space
> it should be a space.
>
> In one case I used perl's extended regex mode to comment a fairly
> hairy regex. This should probably be done a bit more, maybe for all
> of them.
>
> If anybody is so inclined, this could be used as a basis for
> removing the use of bsd indent altogether, as has been suggested
> before, as well as external entab/detab.
Thirteen months after Andrew posted this WIP, I have restructured and
tested this code, and it is now ready to replace the pgindent shell
script as pgindent.pl, attached.
I have tested this version by re-running the 9.1 and 9.2 pgindent runs
and comparing the output, and it is just like Andrew said --- it is the
same, except for the two improvements he mentioned.
A Perl version of pgindent has several advantages:
* more portable; less dependent on utility command variances
* able to run on Windows, assuming someone makes entab and
pg_bsd_indent Windows binaries
* able to fix more limitations of pgindent
I will add documentation about the arguments.
Many thanks to Andrew for his fine work on this. Any objections?
--
Bruce Momjian <[email protected]> http://momjian.us
EnterpriseDB http://enterprisedb.com
+ It's impossible for everything to be true. +
#!/usr/bin/perl
use strict;
use warnings;
use Cwd qw(abs_path getcwd);
use File::Find;
use File::Spec qw(devnull);
use File::Temp;
use IO::Handle;
use Getopt::Long;
use Readonly;
# Update for pg_bsd_indent version
Readonly my $INDENT_VERSION => "1.1";
Readonly my $devnull => File::Spec->devnull;
# Common indent settings
my $indent_opts =
"-bad -bap -bc -bl -d0 -cdb -nce -nfc1 -di12 -i4 -l79 -lp -nip -npro -bbb";
# indent-dependant settings
my $extra_opts = "";
my ($typedefs_file, $code_base, $excludes, $indent, $build);
my %options = (
"typedefs=s" => \$typedefs_file,
"code-base=s" => \$code_base,
"excludes=s" => \$excludes,
"indent=s" => \$indent,
"build" => \$build,);
GetOptions(%options) || die "bad command line";
run_build($code_base) if ($build);
# command line option wins, then first non-option arg,
# then environment (which is how --build sets it) ,
# then locations. based on current dir, then default location
$typedefs_file ||= shift if @ARGV && $ARGV[0] !~ /\\.[ch]$/;
$typedefs_file ||= $ENV{PGTYPEDEFS};
# build mode sets PGINDENT and PGENTAB
$indent ||= $ENV{PGINDENT} || $ENV{INDENT} || "pg_bsd_indent";
my $entab = $ENV{PGENTAB} || "entab";
# no non-option arguments given. so do everything in the current directory
$code_base ||= '.' unless @ARGV;
# if it's the base of a postgres tree, we will exclude the files
# postgres wants excluded
$excludes ||= "$code_base/src/tools/pgindent/exclude_file_patterns"
if $code_base && -f "$code_base/src/tools/pgindent/exclude_file_patterns";
# globals
my @files;
my $filtered_typedefs_fh;
sub check_indent
{
system("entab < $devnull");
if ($?)
{
print STDERR
"Go to the src/tools/entab directory and do 'make' and 'make install'.\n",
"This will put the 'entab' command in your path.\n",
"Then run $0 again.\n";
exit 1;
}
system("$indent -? < $devnull > $devnull 2>&1");
if ($? >> 8 != 1)
{
print STDERR
"You do not appear to have 'indent' installed on your
system.\n";
exit 1;
}
if (`$indent -V` !~ m/ $INDENT_VERSION$/)
{
print STDERR
"You do not appear to have $indent version $INDENT_VERSION installed on your
system.\n";
exit 1;
}
system("$indent -gnu < $devnull > $devnull 2>&1");
if ($? == 0)
{
print STDERR
"You appear to have GNU indent rather than BSD indent.\n",
"See the pgindent/README file for a description of its
problems.\n";
$extra_opts = "-cdb -bli0 -npcs -cli4 -sc";
}
else
{
$extra_opts = "-cli1";
}
}
sub load_typedefs
{
# try fairly hard to find the typedefs file if it's not set
foreach my $try ('.', 'src/tools/pgindent', '/usr/local/etc')
{
$typedefs_file ||= "$try/typedefs.list"
if (-f "$try/typedefs.list");
}
# try to find typedefs by moving up directory levels
my $tdtry = "..";
foreach (1 .. 5)
{
$typedefs_file ||= "$tdtry/src/tools/pgindent/typedefs.list"
if (-f "$tdtry/src/tools/pgindent/typedefs.list");
$tdtry = "$tdtry/..";
}
die "no typedefs file" unless $typedefs_file && -f $typedefs_file;
open(my $typedefs_fh, '<', $typedefs_file) || die "opening
$typedefs_file: $!";
my @typedefs = <$typedefs_fh>;
close($typedefs_fh);
# remove certain entries
@typedefs = grep { ! m/^(FD_SET|date|interval|timestamp|ANY)\n?$/ }
@typedefs;
# write filtered typedefs
my $filter_typedefs_fh = new File::Temp(TEMPLATE => "pgtypedefXXXXX");
print $filter_typedefs_fh @typedefs;
$filter_typedefs_fh->close();
# temp file remains because we return a file handle reference
return $filter_typedefs_fh;
}
sub process_exclude
{
if ($excludes && @files)
{
open(my $eh, '<', $excludes) || die "opening $excludes";
while (my $line = <$eh>)
{
chomp $line;
my $rgx;
eval " \$rgx = qr!$line!;";
@files = grep { $_ !~ /$rgx/ } @files if $rgx;
}
close($eh);
}
}
sub read_source
{
my $source_filename = shift;
my $source;
open(my $src_fd, '<', $source_filename)
|| die "opening $source_filename: $!";
local ($/) = undef;
$source = <$src_fd>;
close($src_fd);
return $source;
}
sub write_source
{
my $source = shift;
my $source_filename = shift;
open(my $src_fh, '>', $source_filename)
|| die "opening $source_filename: $!";
print $src_fh $source;
close($src_fh);
}
sub pre_indent
{
my $source = shift;
# remove trailing whitespace
$source =~ s/\h+$//gm;
## Comments
# Convert // comments to /* */
$source =~ s!^(\h*)//(.*)$!$1/* $2 */!gm;
# 'else' followed by a single-line comment, followed by
# a brace on the next line confuses BSD indent, so we push
# the comment down to the next line, then later pull it
# back up again. Add space before _PGMV or indent will add
# it for us.
# AMD: A symptom of not getting this right is that you see errors like:
# FILE: ../../../src/backend/rewrite/rewriteHandler.c
# Error@2259:
# Stuff missing from end of file
$source =~ s!(\}|\h)else\h*(/\*)(.*\*/)\h*$!$1else\n $2 _PGMV$3!gm;
# Indent multi-line after-'else' comment so BSD indent will move it
# properly. We already moved down single-line comments above.
# Check for '*' to make sure we are not in a single-line comment that
# has other text on the line.
$source =~ s!(\}|\h)else\h*(/\*[^*]*)\h*$!$1else\n $2!gm;
# Mark some comments for special treatment later
$source =~ s!/\* +---!/*---X_X!g;
## Other
# Work around bug where function that defines no local variables
# misindents switch() case lines and line after #else. Do not do
# for struct/enum.
my @srclines = split(/\n/, $source);
foreach my $lno (1 .. $#srclines)
{
my $l2 = $srclines[$lno];
# Line is only a single open brace in column 0
next unless $l2 =~ /^\{\h*$/;
# previous line has a closing paren
next unless $srclines[ $lno - 1 ] =~ /\)/;
# previous line was struct, etc.
next
if $srclines[ $lno - 1 ] =~
m!=|^(struct|enum|\h*typedef|extern\h+"C")!;
$srclines[$lno] = "$l2\nint pgindent_func_no_var_fix;";
}
$source = join("\n", @srclines) . "\n"; # make sure there's a final
\n
# Prevent indenting of code in 'extern "C"' blocks.
# we replace the braces with comments which we'll reverse later
my $extern_c_start = '/* Open extern "C" */';
my $extern_c_stop = '/* Close extern "C" */';
$source =~
s!(^#ifdef\h+__cplusplus.*\nextern\h+"C"\h*\n)\{\h*$!$1$extern_c_start!gm;
$source =~ s!(^#ifdef\h+__cplusplus.*\n)\}\h*$!$1$extern_c_stop!gm;
return $source;
}
sub post_indent
{
my $source = shift;
my $source_filename = shift;
# put back braces for extern "C"
$source =~ s!^/\* Open extern "C" \*/$!{!gm;
$source =~ s!^/\* Close extern "C" \*/$!}!gm;
## Comments
# remove special comment marker
$source =~ s!/\*---X_X!/* ---!g;
# Pull up single-line comment after 'else' that was pulled down above
$source =~ s!else\n\h+/\* _PGMV!else\t/*!g;
# Indent single-line after-'else' comment by only one tab.
$source =~ s!(\}|\h)else\h+(/\*.*\*/)\h*$!$1else\t$2!gm;
# Add tab before comments with no whitespace before them (on a tab stop)
$source =~ s!(\S)(/\*.*\*/)$!$1\t$2!gm;
# Remove blank line between opening brace and block comment.
$source =~ s!(\t*\{\n)\n(\h+/\*)$!$1$2!gm;
# cpp conditionals
# Reduce whitespace between #endif and comments to one tab
$source =~ s!^\#endif\h+/\*!#endif /*!gm;
# Remove blank line(s) before #else, #elif, and #endif
$source =~ s!\n\n+(\#else|\#elif|\#endif)!\n$1!g;
# Add blank line before #endif if it is the last line in the file
$source =~ s!\n(#endif.*)\n\z!\n\n$1\n!;
## Functions
# Work around misindenting of function with no variables defined.
$source =~ s!^\h*int\h+pgindent_func_no_var_fix;\h*\n{1,2}!!gm;
# Use a single space before '*' in function return types
$source =~ s!^([A-Za-z_]\S*)\h+\*$!$1 *!gm;
# Move prototype names to the same line as return type. Useful
# for ctags. Indent should do this, but it does not. It formats
# prototypes just like real functions.
my $ident = qr/[a-zA-Z_][a-zA-Z_0-9]*/;
my $comment = qr!/\*.*\*/!;
$source =~ s!
(\n$ident[^(\n]*)\n # e.g. static void
(
$ident\(\n? # func_name(
(.*,(\h*$comment)?\n)* # args b4
final ln
.*\);(\h*$comment)?$ # final line
)
!$1 . (substr($1,-1,1) eq '*' ? '' : ' ') . $2!gmxe;
## Other
# Remove too much indenting after closing brace.
$source =~ s!^\}\t\h+!}\t!gm;
# Workaround indent bug that places excessive space before 'static'.
$source =~ s!^static\h+!static !gm;
# Remove leading whitespace from typedefs
$source =~ s!^\h+typedef enum!typedef enum!gm
if $source_filename =~ 'libpq-(fe|events).h$';
# Remove trailing blank lines
$source =~ s!\n+\z!\n!;
return $source;
}
sub run_indent
{
my $source = shift;
my $error_message = shift;
my $cmd = "$indent $indent_opts $extra_opts -U" .
$filtered_typedefs_fh->filename;
my $tmp_fh = new File::Temp(TEMPLATE => "pgsrcXXXXX");
my $filename = $tmp_fh->filename;
print $tmp_fh $source;
$tmp_fh->close();
$$error_message = `$cmd $filename 2>&1`;
return "" if ($? || length($$error_message) > 0);
unlink "$filename.BAK";
open(my $src_out, '<', $filename);
local ($/) = undef;
$source = <$src_out>;
close($src_out);
return $source;
}
# XXX Ideally we'd implement entab/detab in pure perl.
sub detab
{
my $source = shift;
my $tmp_fh = new File::Temp(TEMPLATE => "pgdetXXXXX");
print $tmp_fh $source;
$tmp_fh->close();
open(my $entab, '-|', "$entab -d -t4 -qc " . $tmp_fh->filename);
local ($/) = undef;
$source = <$entab>;
close($entab);
return $source;
}
sub entab
{
my $source = shift;
my $tmp_fh = new File::Temp(TEMPLATE => "pgentXXXXX");
print $tmp_fh $source;
$tmp_fh->close();
open(my $entab, '-|',
"$entab -d -t8 -qc " . $tmp_fh->filename . " | $entab -t4 -qc");
local ($/) = undef;
$source = <$entab>;
close($entab);
return $source;
}
# for development diagnostics
sub diff
{
my $pre = shift;
my $post = shift;
my $flags = shift || "";
print STDERR "running diff\n";
my $pre_fh = new File::Temp(TEMPLATE => "pgdiffbXXXXX");
my $post_fh = new File::Temp(TEMPLATE => "pgdiffaXXXXX");
print $pre_fh $pre;
print $post_fh $post;
$pre_fh->close();
$post_fh->close();
system("diff $flags " . $pre_fh->filename . " " .
$post_fh->filename . " >&2");
}
sub run_build
{
eval "use LWP::Simple;";
my $code_base = shift || '.';
my $save_dir = getcwd();
# look for the code root
foreach (1 .. 5)
{
last if -d "$code_base/src/tools/pgindent";
$code_base = "$code_base/..";
}
die "no src/tools/pgindent directory in $code_base"
unless -d "$code_base/src/tools/pgindent";
chdir "$code_base/src/tools/pgindent";
my $rv = getstore("http://buildfarm.postgresql.org/cgi-bin/typedefs.pl";,
"tmp_typedefs.list");
die "fetching typedefs.list" unless is_success($rv);
$ENV{PGTYPEDEFS} = abs_path('tmp_typedefs.list');
$rv =
getstore("ftp://ftp.postgresql.org/pub/dev/indent.netbsd.patched.tgz";,
"indent.netbsd.patched.tgz");
die "fetching indent.netbsd.patched.tgz" unless is_success($rv);
# XXX add error checking here
mkdir "bsdindent";
chdir "bsdindent";
system("tar -z -xf ../indent.netbsd.patched.tgz");
system("make > $devnull 2>&1");
$ENV{PGINDENT} = abs_path('indent');
chdir "../../entab";
system("make > $devnull 2>&1");
$ENV{PGENTAB} = abs_path('entab');
chdir $save_dir;
}
sub build_clean
{
my $code_base = shift || '.';
# look for the code root
foreach (1 .. 5)
{
last if -d "$code_base/src/tools/pgindent";
$code_base = "$code_base/..";
}
die "no src/tools/pgindent directory in $code_base"
unless -d "$code_base/src/tools/pgindent";
chdir "$code_base";
system("rm -rf src/tools/pgindent/bsdindent");
system("git clean -q -f src/tools/entab src/tools/pgindent");
}
# main
# get the list of files under code base, if it's set
File::Find::find(
{ wanted => sub {
my ($dev, $ino, $mode, $nlink, $uid, $gid);
(($dev, $ino, $mode, $nlink, $uid, $gid) = lstat($_))
&& -f _
&& /^.*\.[ch]\z/s
&& push(@files, $File::Find::name);
}
},
$code_base) if $code_base;
process_exclude();
$filtered_typedefs_fh = load_typedefs();
check_indent();
# make sure we process any non-option arguments.
push(@files, @ARGV);
foreach my $source_filename (@files)
{
my $source = read_source($source_filename);
my $error_message = '';
$source = pre_indent($source);
# Protect backslashes in DATA() and wrapping in CATALOG()
$source = detab($source);
$source =~ s!^((DATA|CATALOG)\(.*)$!/*$1*/!gm;
$source = run_indent($source, \$error_message);
if ($source eq "")
{
print STDERR "Failure in $source_filename: " . $error_message .
"\n";
next;
}
# Restore DATA/CATALOG lines; must be done here so tab alignment is
preserved
$source =~ s!^/\*((DATA|CATALOG)\(.*)\*/$!$1!gm;
$source = entab($source);
$source = post_indent($source, $source_filename);
write_source($source, $source_filename);
}
build_clean($code_base) if $build;
--
Sent via pgsql-hackers mailing list ([email protected])
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers
