Advanced-Config
view release on metacpan or search on metacpan
lib/Advanced/Config/Reader.pm view on Meta::CPAN
my %hide_section;
while ( <$READ_CONFIG> ) {
chomp;
my $line = $_; # Save so can use in fish logging later on.
my ($tv, $ln, $cmt, $lq, $rq) = parse_line ( $line, $opts );
if ( $ln eq "" ) {
DBUG_PRINT ("READ", "READ LINE: %s", $line);
next; # Skip to the next line if only comments found.
}
# Check for lines with no tag/value pairs in them ...
if ( ! $tv ) {
DBUG_PRINT ("READ", "READ LINE: %s", $line);
# EX: . ${file} --- Sourcing in ${file} ...
if ( $ln =~ m/^${src_str}\s+(.+)$/i ) {
my $src = $1;
my $def_section = "";
if ( $cmt =~ m/(^|${lbl_sep})${sect_str}(${lbl_sep}|$)/ ) {
$def_section = $section;
}
my $res = source_file ( $cfg, $def_section, $src, $file );
return DBUG_RETURN (0) unless ( $res );
next;
}
# EX: [ ${section} ] --- Starting a new section ...
if ( $ln =~ m/^${lb}\s*(.+?)\s*${rb}$/ ) {
$section = make_new_section ( $cfg, $1 );
$hide_section{$section} = 0; # Assume not sensitive ...
if ( $cmt =~ m/(^|${lbl_sep})${hide_str}(${lbl_sep}|$)/ ||
should_we_hide_sensitive_data ( $section ) ) {
$hide_section{$section} = 1;
}
next;
}
# Don't know what the config file was thinking of ...
# Don't bother expanding any variables encountered.
DBUG_PRINT ("error", "<Previous line ignored. Unknown format!>");
next;
}
# ------------------------------------------------------------------
# If you get here, you know it's a tag/value pair to parse ...
# Don't forget that any comment can include processing instructions!
# ------------------------------------------------------------------
# Go to the requested section ...
$cfg = $pcfg->get_section ( $section, 1 );
my ($tag, $value, $prefix, $t2) = _split_assign ( $opts, $ln );
# Don't export individually if doing a batch export ...
# If the export option is used, invert the meaning ...
my $export_flag = 0; # Assume not exporting this tag to %ENV ...
if ( $prefix ) {
$export_flag = $opts->{export} ? 0 : 1;
} elsif ( $cmt =~ m/(^|${lbl_sep})${export_str}(${lbl_sep}|$)/ ) {
$export_flag = $opts->{export} ? 0 : 1;
}
# Is the line info sensitive & should it be hidden/masked in fish ???
my $hide = 0;
if ( $hide_section{$section} ||
$cmt =~ m/(^|${lbl_sep})${encrypt_str}(${lbl_sep}|$)/ ||
$cmt =~ m/(^|${lbl_sep})${hide_str}(${lbl_sep}|$)/ ||
should_we_hide_sensitive_data ( $tag, 1 ) ) {
$hide = 1 unless ( $opts->{dbug_test_use_case_hide_override} );
}
if ( $hide ) {
# Some random length so we can't assume the value from the mask used!
my $mask = "*"x8;
if ( $value eq "" ) {
if ( is_assign_spaces ( $opts ) ) {
$line =~ s/^(\s*\S+\s+)/${1}${mask} /;
} else {
$line =~ s/(\s*${assign_str})\s*/${1} ${mask} /;
}
} else {
my $hide_value = convert_to_regexp_string ( $value, 1 );
if ( is_assign_spaces ( $opts ) ) {
$line =~ s/^(\s*\S+\s+)${hide_value}/${1}${mask}/;
} else {
$line =~ s/(\s*${assign_str}\s*)${hide_value}/${1}${mask}/;
}
}
} elsif ( $cmt =~ m/(^|${lbl_sep})${decrypt_str}(${lbl_sep}|$)/ ) {
# Don't hide the line in fish, but hide it's value processing ...
$hide = 1 unless ( $opts->{dbug_test_use_case_hide_override} );
}
DBUG_PRINT ("READ", "READ LINE: %s", $line);
# Remove any balanced quotes ... (must do after hide)
$value =~ s/^${lq}(.*)${rq}$/$1/ if ( $lq );
if ( $tag =~ m/^(shft3+)$/i ) {
my $m = "You can't override special variable '${1}'."
. " Ignoring this line in the config file.\n";
if ( $skip_warns_due_to_make_test ) {
DBUG_PRINT ("WARN", $m);
} else {
warn $m;
}
next;
}
# Was the tag's value encryped?? Then we need to decrypt it ...
my $still_encrypted = 0;
if ( $cmt =~ m/(^|${lbl_sep})${decrypt_str}(${lbl_sep}|$)/ ) {
$value = _reverse_escape_sequences ( $value, $opts );
if ( $opts->{disable_decryption} ) {
$still_encrypted = 1; # Doesn't get decrypted.
} else {
$value = decrypt_value ( $value, $t2, $opts, $file );
}
}
# See if we can expand variables in $value ???
my $still_variables = 0;
if ( $opts->{disable_variables} ) {
$still_variables = ( $value =~ m/${lv}.+${rv}/ ) ? 1 : 0;
} elsif ( ! $still_encrypted ) {
($value, $hide) = expand_variables ( $cfg, $value, $file, $hide, ($lq ? 0 : 1) );
if ( $hide == -1 ) {
# $still_encrypted = $still_variables = 1;
$still_variables = 1; # Variable(s) points to encrypted data.
}
}
# Export one value to %ENV ... (once set, can't back it out again!)
$cfg->export_tag_value_to_ENV ( $tag, $value, $hide ) if ($export_flag);
# Add to the current section in the Advanced::Config object ...
$cfg->_base_set ($tag, $value, $file, $hide, $still_encrypted, $still_variables);
} # End while reading the config file ...
close ( $READ_CONFIG );
DBUG_RETURN (1);
}
# ==============================================================
=item $boolean = source_file ($config, $def_sct, $new_file, $curr_file)
This is a private method called by I<read_config> to source in the requested
config file and merge the results into the current config file.
If I<$def_sct> is given, it will be the name of the current section that the
sourced in file is to use for it's default unlabeled section. If the default
section name has been hard coded in the config file, this value overrides it.
The I<$new_file> may contain variables and after they are expanded the
source callback function is called before I<load_config()> is called.
See L<Advanced::Config::lookup_one_variable> for rules on variable expansion.
If I<$new_file> is a relative path, it's a relative path from the location
of I<$curr_file>, not the program's current directory!
If a source callback was set up, it will call it here.
This method will also handle the removal of decryption related options if new
ones weren't provided by the callback function. See Advanced::Config::Options
for more details.
Returns B<1> if the new file successfully loaded. Else B<0> if something went
wrong during the load!
=cut
sub source_file
{
DBUG_ENTER_FUNC (@_);
my $cfg = shift;
my $defaultSection = shift; # The new default section if not "".
my $new_file = shift; # May contain variables to expand ...
my $old_file = shift; # File we're currently parsing. (has abs path)
my $rOpts = $cfg->get_cfg_settings (); # The Read Options ...
local $global_sections{OVERRIDE} = $defaultSection if ( $defaultSection );
my $pcfg = $cfg->get_section (); # Back to the main/default section ...
my $file = $new_file = expand_variables ($pcfg, $new_file, undef, undef, 1);
# Get the full name of the file we're sourcing in ...
$file = $pcfg->_fix_path ( $file, dirname ( $old_file ) );
unless ( -f $file && -r _ ) {
lib/Advanced/Config/Reader.pm view on Meta::CPAN
DBUG_RETURN ($value);
}
# ==============================================================
=item @ret[0..4] = parse_line ( $input, \%opts )
This is a private method called by I<read_config> to parse each line of the
config file as it's read in. It's main purpose is to strip off leading and
trailing spaces and any comments it might find on the input line. It also
tells if the I<$input> contains a tag/value pair.
It returns 5 values: ($tv_flg, $line, $comment, $lQuote, $rQuote)
B<$tv_flg> - True if I<$line> contains a tag/value pair in it, else false.
B<$line> - The trimmed I<$input> line minus any comments.
B<$comment> - The comment stripped out of the original input line minus the
leading comment symbol(s).
B<$lQuote> & B<rQuote> - Only set if I<$tv_flg> is true and I<$lQuote> was
the 1st char of the value and I<$rQuote> was the last char of the tag's value.
If the value wasn't surrounded by balanced quotes, both return values will be
the empty string B<"">.
If these quotes are returned, it expects the caller to remove them from the
tag's value. The returned values for these quote chars are suitable for use as
is in a RegExpr. The caller must do this in order to preserve potential
leading/trailing spaces.
=cut
sub parse_line
{
DBUG_MASK_NEXT_FUNC_CALL (0); # Masks ${line}!
DBUG_ENTER_FUNC ( @_ );
my $line = shift;
my $opts = (ref ($_[0]) eq "HASH" ) ? $_[0] : {@_};
# Mask the ${line} return value in fish ...
# Only gets unmasked in the test scripts: t/*.t.
# Always pause since by the time we detect if it should be
# hidden or not it's too late. We've already written it to fish!
unless ( $opts->{dbug_test_use_case_parse_override} ) {
DBUG_MASK ( 1 );
DBUG_PAUSE ();
}
# Strip of leading & trailing spaces ...
$line =~ s/^\s+//;
$line =~ s/\s+$//;
my $default_quotes = using_default_quotes ( $opts );
my $comment = convert_to_regexp_string ($opts->{comment}, 1);
my ($tag, $value) = _split_assign ( $opts, $line, 1 );
my ($l_quote, $r_quote, $tv_pair_flag) = ("", "", 0);
my $var_line = $line;
unless ( defined $tag && defined $value ) {
$tag = $value = undef; # It's not a tag/value pair ...
} elsif ( $tag eq "" || $tag =~ m/${comment}/ ) {
$tag = $value = undef; # It's not a valid tag ...
} else {
# It looks like a tag/value pair to me ...
$tv_pair_flag = 1;
if ( $opts->{disable_quotes} ) {
; # Don't do anything ...
} elsif ( $default_quotes ) {
if ( $value =~ m/^(['"])/ ) {
$l_quote = $r_quote = $1; # A ' or ". (Never both)
}
# User defined quotes ...
} else {
my $q = convert_to_regexp_string ($opts->{quote_left}, 1);
if ( $value =~ m/^(${q})/ ) {
$l_quote = $q;
$r_quote = convert_to_regexp_string ($opts->{quote_right}, 1);
}
}
$var_line = $value;
}
# Comment still in value, but still haven't proved any quotes are balanced.
DBUG_PRINT ("DEBUG", "Tag (%s), Value (%s), Proposed Left (%s), Right (%s)",
$tag, $value, $l_quote, $r_quote);
my $cmts = "";
# Was the value in the tag/value pair starting with a left quote?
if ( $tv_pair_flag && $l_quote ne "" ) {
my ($q1, $val2, $q2);
# Now check if they were balanced ...
if ( $value =~ m/^(${l_quote})(.*)(${r_quote})(\s*${comment}.*$)/ ) {
($q1, $val2, $q2, $cmts) = ($1, $2, $3, $4); # Has a comment ...
} elsif ( $value =~ m/^(${l_quote})(.*)(${r_quote})\s*$/ ) {
($q1, $val2, $q2, $cmts) = ($1, $2, $3, ""); # Has no comment ...
}
# If balanced quotes were found ...
if ( $q1 ) {
# If the surrounding quotes don't have quotes inside them ...
# IE not malformed ...
unless ( $val2 =~ m/${l_quote}/ || $val2 =~ m/${r_quote}/ ) {
my $cmt2 = convert_to_regexp_string ($cmts);
$cmts =~ s/^\s*${comment}\s*//; # Remove comment symbol ...
$line =~ s/${cmt2}$// if ($cmt2 ne "" ); # Remove the comments ...
DBUG_PRINT ("LINE", "Balanced Quotes encountered for removal ...");
return DBUG_RETURN ( $tv_pair_flag, $line, $cmts, $l_quote, $r_quote);
}
}
}
# The Quotes weren't balanced, so they can no longer be removed from
# arround the value of what's returned!
$l_quote = $r_quote = "";
# ----------------------------------------------------------------------
# If no comments in the line, just return the trimmed string ...
# Both tests are needed due to custom comment/assign strings!
# ----------------------------------------------------------------------
if ( $line !~ m/${comment}/ ) {
DBUG_PRINT ("LINE", "Simply no comments to worry about ...");
return DBUG_RETURN ( $tv_pair_flag, $line, "", "", "" );
}
# Handles case where a comment char embedded in the assignment string.
if ( $tv_pair_flag && $value !~ m/${comment}/ ) {
DBUG_PRINT ("LINE", "Simply no comments in the value to worry about ...");
return DBUG_RETURN ( $tv_pair_flag, $line, "", "", "" );
}
# ----------------------------------------------------------------------
# If not protected by balanced quotes, verify the comment symbol detected
# isn't actually a variable modifier. Variables are allowed in most places
# in the config file, not just in tag/value pairs.
# ----------------------------------------------------------------------
# The left & right anchor points for variable substitution ...
my $lvar = convert_to_regexp_string ($opts->{variable_left}, 1);
my $rvar = convert_to_regexp_string ($opts->{variable_right}, 1);
# Determine what value to use in variable substitutions that doesn't include
# a variable tag, or a comment tag, or a value in the $line.
my $has_no_cmt;
foreach ("A" .. "Z", "@") {
$has_no_cmt = ${_}x10;
last unless ( $has_no_cmt =~ m/${comment}/ ||
$has_no_cmt =~ m/${lvar}/ ||
$has_no_cmt =~ m/${rvar}/ ||
$line =~ m/${has_no_cmt}/ );
}
if ( $has_no_cmt eq "@"x10 ) {
warn ("May be having variable substitution issues in parse_line()!\n");
}
# Strip out all the variables from the value ...
# Assumes processing variables from left to right ...
# Need to evaluate even if variables are disabled to parse correctly ...
my @parts = parse_for_variables ($var_line, 1, $opts);
my $cmt_found = 0;
my $count_var = 0;
my @data;
while (defined $parts[0]) {
$cmt_found = $parts[3];
push (@data, $var_line);
last if ($cmt_found);
$var_line = $parts[0] . $has_no_cmt . $parts[2];
@parts = parse_for_variables ($var_line, 1, $opts);
++$count_var;
}
push (@data, $var_line);
my $unbalanced_leading_var_anchor_with_comments = 0;
if ( $cmt_found && $parts[0] =~ m/(\s*${comment}\s*)(.*$)/ ) {
# parts[1] is parts[7] trimmed ... so join back together with untrimmed.
$cmts = $2 . $opts->{variable_left} . $parts[7]
. $opts->{variable_right} . $parts[2];
my $str = convert_to_regexp_string ( $1 . $cmts );
$line =~ s/${str}$//;
DBUG_PRINT ("LINE", "Variables encountered with variables in comment ...");
return DBUG_RETURN ( $tv_pair_flag, $line, $cmts, "", "");
} elsif ( $count_var ) {
if ( $var_line =~ m/(\s*${comment}\s*)(.*)$/ ) {
$cmts = $2;
if ( $cmts =~ m/${has_no_cmt}/ ) {
$unbalanced_leading_var_anchor_with_comments = 1;
} else {
my $cmt2 = convert_to_regexp_string ($1 . $cmts);
$line =~ s/${cmt2}$//;
DBUG_PRINT ("LINE", "Variables encountered with constant comment ...");
}
} else {
$cmts = "";
DBUG_PRINT ("LINE", "Variables encountered without comments ...");
}
unless ( $unbalanced_leading_var_anchor_with_comments ) {
return DBUG_RETURN ( $tv_pair_flag, $line, $cmts, "", "");
}
}
# ---------------------------------------------------------------------------
# Corrupted variable definition with variables in the comments ...
# Boy things are getting difficult to parse. Reverse the previous variable
# substitutions until the all variables in the comments are unexpanded again!
# Does a greedy RegExp to grab the 1st comment string encountered.
# ---------------------------------------------------------------------------
if ( $unbalanced_leading_var_anchor_with_comments ) {
$cmts = "";
foreach my $l (reverse @data) {
if ( $l =~ m/\s*${comment}\s*(.*)$/ ) {
$cmts = $1;
last unless ( $cmts =~ m/${has_no_cmt}/ );
$cmts = "";
}
}
if ( $cmts ne "" ) {
my $cmt2 = convert_to_regexp_string ($cmts);
$line =~ s/\s*${comment}\s*${cmt2}$//;
DBUG_PRINT ("LINE", "Unbalanced var def encountered with var comments ...");
return DBUG_RETURN ( $tv_pair_flag, $line, $cmts, "", "");
}
# If you get here, assume it's not a tag/value pair even if it is!
# I know I can no longer hope to parse it correctly without a test case.
# But I really don't think it's possible to get here anymore ...
warn ("Corrupted variable definition encountered. Can't split out the comment with variables in it correctly!\n");
return DBUG_RETURN ( 0, $line, "", "", "");
}
# ----------------------------------------------------------------------
# No variables, no balanced quotes ...
# But I still think there's a comment to remove!
# ----------------------------------------------------------------------
if ( $tv_pair_flag && $value =~ m/(\s*${comment}\s*)(.*)$/ ) {
$cmts = $2;
my $cmt2 = convert_to_regexp_string ($1 . $cmts);
$line =~ s/${cmt2}$//; # Remove the comment from the line.
DBUG_PRINT ("LINE", "Last ditch effort to remove the comment from the value ...");
return DBUG_RETURN ( $tv_pair_flag, $line, $cmts, "", "");
}
$cmts = $line;
$line =~ s/\s*${comment}.*$//; # Strip off any comments ....
$cmts = substr ( $cmts, length ($line) ); # Grab the comments ...
$cmts =~ s/^\s*${comment}\s*//; # Remove comment symbol ...
DBUG_PRINT ("LINE", "Last ditch effort to remove the comment from the line ...");
DBUG_RETURN ( $tv_pair_flag, $line, $cmts, "", "");
}
# ==============================================================
=item ($v[, $h]) = expand_variables ( $config, $string[, $file[, $sensitive[, trim]]] )
This function takes the provided I<$string> and expands any embedded variables
in this string similar to how it's handled by a Unix shell script.
The optional I<$file> tells which file the string was read in from.
The optional I<$sensitive> when set to a non-zero value is used to disable
B<fish> logging when it's turned on because the I<$string> being passed contains
sensitive information.
The optional I<$trim> tells if you may trim the results before it's returned.
It returns the new value $v, once all the variable substitution(s) have
occurred. And optionally a second return value $h that tells if B<fish> was
paused during the expansion of that value due to something being sensitive.
This 2nd return value $h is meaningless in most situations, so don't ask for it.
All variables are defined as B<${>I<...>B<}>, where I<...> is the variable you
wish to substitute. If something isn't surrounded by a B<${> + B<}> pair, it's
not a variable.
A config file exampe:
tmp1 = /tmp/work-1
tmp2 = /tmp/work-2
opt = 1
date = 2011-02-03
logs = ${tmp${opt}}/log-${date}.txt
date = 2012-12-13
So when passed "${tmp${opt}}/log-${date}.txt", it would return:
/tmp/work-1/log-2011-02-03.txt
And assigned it to B<logs>.
As you can see multiple variable substitutions may be expanded in a single
string as well as nested substitutions. And when the variable substitution is
done while reading in the config file, all the values used were defined before
the tag was referenced.
Should you call this method after the config file was loaded you get slightly
different results. In that case the final tag value is used instead and the
2nd date in the above example would have been used in it's place.
See L<Advanced::Config::lookup_one_variable> for more details on how it
evaluates individual variables.
As a final note, if one or more of the referenced variables holds encrypted
values that haven't yet been decrypted, those variables are not resolved. But
all variables that don't contain encrypted data are resolved.
=cut
# ==============================================================
sub expand_variables
{
my $config = shift; # For the current section of config obj ...
my $value = shift; # The value to parse for variables ...
my $file = shift || ""; # The config file the value came from ...
my $mask_flag = shift || 0; # Hide/mask sensitive info written to fish?
my $trim_flag = shift || 0; # Tells if we should trim the result or not.
# Only mask ${value} if ${mask_flag} is true ...
DBUG_MASK_NEXT_FUNC_CALL (1) if ( $mask_flag );
DBUG_ENTER_FUNC ( $config, $value, $file, $mask_flag, $trim_flag, @_);
my $opts = $config->get_cfg_settings (); # The Read Options ...
my $pcfg = $config->get_section(); # Get the main/parent section to work with!
# Don't write to Fish if we're hiding any values ...
if ( $mask_flag ) {
DBUG_PAUSE ();
DBUG_MASK ( 0 );
}
# The 1st split of the value into it's component parts ...
my ($left, $tag, $right, $cmt_flag, $mod_tag, $mod_opt, $mod_val, $ot) =
parse_for_variables ( $value, 0, $opts );
# Any variables to substitute ???
unless ( defined $tag ) {
return DBUG_RETURN ( $value, $mask_flag ); # nope ...
}
my $output = $value;
my %encrypt_vars;
my $encrypt_cnt = 0;
my $encrypt_fmt = "_"x50 . "ENCRYPT_%02d" . "-"x50;
my ($lv, $rv) = ( convert_to_regexp_string ($opts->{variable_left}),
convert_to_regexp_string ($opts->{variable_right}) );
# While there are still variables to process ...
while ( defined $tag ) {
my ( $val, $mask );
my $do_mod_lookup = 0; # Very rarely set to true ...
# ${tag} and ${mod_tag} will never have the same value ...
# ${mod_tag} will amost always be undefinded.
# If both are defined, we'll almost always end up using ${mod_tag} as
# the real variable to expand! But we check to be sure 1st.
( $val, $mask ) = $config->lookup_one_variable ( $tag );
# It's extreemly rare to have this "if statement" evalate to true ...
if ( (! defined $val) && defined $mod_tag ) {
( $val, $mask ) = $config->lookup_one_variable ( $mod_tag );
# -----------------------------------------------------------------
# If we're using variable modifiers, it doesn't matter if the
# varible exists or not. The modifier gets evaluated!
# So checking if the undefined $mod_tag needs to be masked or not ...
# -----------------------------------------------------------------
unless ( defined $val ) {
$mask = should_we_hide_sensitive_data ( $mod_tag );
}
$do_mod_lookup = 1; # Yes, apply the modifiers!
}
# Use a place holder if the variable references data that is still encrypted.
if ( $mask == -1 ) {
$mask_flag = -1;
$val = sprintf ($encrypt_fmt, ++$encrypt_cnt);
# If the place holder contains variable anchors abort the substitutions.
last if ( $val =~ m/${lv}/ || $val =~ m/${rv}/ );
$encrypt_vars{$val} = $tag;
$do_mod_lookup = 0;
}
# Doing some accounting to make sure any sensitive data doesn't
# show up in the fish logs from now on.
if ( $mask && ! $mask_flag ) {
$mask_flag = 1;
DBUG_PAUSE ();
DBUG_MASK ( 0 );
}
if ( $do_mod_lookup ) {
my $m;
($val, $m) = apply_modifier ( $config, $val, $mod_tag, $mod_opt, $mod_val, $file );
if ( $m == -2 ) {
# The name of the variable changed & points to an encrypted value.
$val = $opts->{variable_left} . ${val} . $opts->{variable_right};
} elsif ( $m && ! $mask_flag ) {
$mask_flag = 1;
DBUG_PAUSE ();
DBUG_MASK ( 0 );
}
}
# Rebuild the output string so we can look for more variables ...
if ( defined $val ) {
$output = $left . $val . $right;
} else {
$output = $left . $right;
}
# Get the next variable to evaluate ...
($left, $tag, $right, $cmt_flag, $mod_tag, $mod_opt, $mod_val, $ot) =
parse_for_variables ( $output, 0, $opts );
} # End while ( defined $tag ) loop ...
# Restore all place holders back into the output string with the
# proper variable name. Have to assume still sensitive even if
# all the placeholders drop out. Since can't tell what else may
# have triggered it.
if ( $mask_flag == -1 ) {
$mask_flag = 1; # Mark sensitive ...
foreach ( keys %encrypt_vars ) {
my $val = $opts->{variable_left} . $encrypt_vars{$_} . $opts->{variable_right};
$mask_flag = -1 if ( $output =~ s/$_/$val/ );
}
}
# Did the variable substitution result in the need to trim things?
if ( $trim_flag ) {
$output =~ s/^\s+//;
$output =~ s/\s+$//;
}
DBUG_RETURN ( $output, $mask_flag );
}
# ==============================================================
=item ($v[, $s]) = apply_modifier ( $config, $value, $tag, $rule, $sub_rule, $file )
This is a helper method to F<expand_variables>. Not for public use.
This function takes the rule specified by I<$rule> and applies it against
the I<$value> with assistance from the I<$sub_rule>.
It returns the edited I<value> and whether applying the modifier made it
I<sensitive>. (-1 means it's an encrypted value. -2 means it's the variable
name that resolves to an encrypted value. 0 - Non-sensitive, 1 - Sensitive.)
See L<https://web.archive.org/web/20200309072646/https://wiki.bash-hackers.org/syntax/pe>
for information on how this can work. This module supports most of the
parameter expansions listed there except for those dealing with arrays. Other
modifier rules may be added upon request.
=cut
# NOTE1: Fish has already been paused if $tag is sensitive. Since this method
# has no idea if the current tag is sensitive or not.
# NOTE2: But still need to mask the return value if referencing sensitive data
# in case the original $tag wasn't sensitive. So in most cases it will
# return not-sensitive even if fish has already been paused!
#
# NOTE3: If sensitive/mask is -1, it's sensitive and not decrypted. In this
# case the returned value is the tag's name, not it's value!
sub apply_modifier
{
DBUG_ENTER_FUNC ( @_ );
my $cfg = shift;
my $value = shift; # The value for ${mod_tag} ...
my $mod_tag = shift; # The tag to apply the rule against!
my $mod_opt = shift; # The rule ...
my $mod_val = shift; # The sub-rule ...
my $file = shift; # The file the tag's from.
my $alt_val = (defined $value) ? $value : "";
# The values to return ...
my $output;
# Values: 0 - Normal non-sensitive return value (99.9% of the time)
# 1 - Sensitive return value.
# -1 - Return value is encrypted.
# -2 - Return value is variable name of encrypted value.
my $mask = 0;
# If looking for a default value ...
if ( ( $mod_opt eq ":+" && $alt_val ne "" ) ||
( $mod_opt =~ m/^:[-=?]$/ && $alt_val eq "" ) ||
( $mod_opt eq "+" && defined $value ) ||
( $mod_opt =~ m/^[-=?]$/ && ! defined $value ) ) {
$output = $mod_val; # Now uses this value as it's default!
if ( $mod_opt eq ":=" || $mod_opt eq "=" ) {
# The variable either doesn't exist or it resolved to "".
# This variant rule says to also set the variable to this value!
$cfg->_base_set ( $mod_tag, $output, $file );
} elsif ( $mod_opt eq ":?" || $mod_opt eq "?" ) {
# In shell scripts, ":?" would cause your script to die with the
# default value as the error message if your var had no value.
# Repeating that logic here.
my $msg = "Encounterd undefined variable ($mod_tag) using shell modifier ${mod_opt}";
$msg .= " in config file: " . basename ($file) if ( $file ne "" );
DBUG_PRINT ("MOD", $msg);
die ( basename ($0) . ": ${mod_tag}: ${output}.\n" );
}
DBUG_PRINT ("MOD",
"The modifier (%s) is overriding the variable with a default value!",
$mod_opt);
# Sub-string removal ...
} elsif ( $mod_opt eq "##" || $mod_opt eq "#" || # From beginning
$mod_opt eq "%%" || $mod_opt eq "%" ) { # From end
my $greedy = ( $mod_opt eq "##" || $mod_opt eq "%%" );
my $leading = ( $mod_opt eq "#" || $mod_opt eq "##" );
my $reverse_msg = ""; # Both the message & control flag ...
$output = $alt_val;
# Now replace shell script wildcards with their Perl equivalents.
# A RegExp can't do non-greedy replaces anchored to the end of string!
# So we need the reverse logic to do so.
my $regExpVal = convert_to_regexp_modifier ($mod_val);
$regExpVal =~ s/[?]/./g; # ? --> . (any one char)
if ( $greedy ) {
$regExpVal =~ s/[*]/.*/g; # * --> .* (zero or more greedy chars)
} elsif ( $leading ) {
$regExpVal =~ s/[*]/(.*?)/g; # * --> (.*?) (zero or more chars)
} elsif ( $regExpVal =~ m/[*]/ ) {
# Non-Greedy with one or more wild cards present ("*")!
$leading = 1; # Was false before.
$regExpVal = reverse ($regExpVal);
$regExpVal =~ s/[*]/(.*?)/g; # * --> (.*?) (zero or more chars)
$output = reverse ($output);
$reverse_msg = " Reversed for non-greedy strip.";
}
if ( $leading ) {
$regExpVal = '^' . $regExpVal;
} else {
# Either greedy trailing or no *'s in trailing regular expression!
$regExpVal .= '$';
}
$output =~ s/${regExpVal}//; # Strip off the matching values ...
$output = reverse ($output) if ( $reverse_msg ne "" );
DBUG_PRINT ("MOD",
"The modifier (%s) converted \"%s\" to \"%s\".%s\nTo trim the value to: %s",
$mod_opt, $mod_val, $regExpVal, $reverse_msg, $output);
} elsif ( $mod_opt eq "LENGTH" ) {
$output = length ( $alt_val );
DBUG_PRINT ("MOD", "Setting the length of variable \${#%s} to: %d.",
$mod_tag, $output);
} elsif ( $mod_opt eq "LIST" ) {
my @lst = $cfg->_find_variables ( $mod_val );
$output = join (" ", @lst);
DBUG_PRINT ("MOD", "Getting all varriables starting with %s", $mod_val);
} elsif ( $mod_opt eq "!" ) {
($output, $mask) = $cfg->lookup_one_variable ( $alt_val );
if ( $mask == -1 ) {
$mask = -2; # Indirect reference to encrypted value
$output = $alt_val; # Replace with new variable name
} elsif ( $mask ) {
DBUG_MASK (0);
}
DBUG_PRINT ("MOD", "Indirectly referencing variable %s (%s)", $alt_val, $mask);
} elsif ( $mod_opt eq "//" ) {
my ($ptrn, $val) = split ("/", $mod_val);
$output = $alt_val;
$output =~ s/${ptrn}/${val}/g;
DBUG_PRINT ("MOD", "Global replacement in %s", $alt_val);
} elsif ( $mod_opt eq "/" ) {
my ($ptrn, $val) = split ("/", $mod_val);
$output = $alt_val;
$output =~ s/${ptrn}/${val}/;
DBUG_PRINT ("MOD", "1st replacement in %s", $alt_val);
} elsif ( $mod_opt eq ":" ) {
my ($offset, $length) = split (":", $mod_val);
if ( defined $length && $length ne "" ) {
$output = substr ( $alt_val, $offset, $length);
} else {
$output = substr ( $alt_val, $offset);
}
DBUG_PRINT ("MOD", "Substring (%s)", $output);
# The 6 case manipulation modifiers ...
} elsif ( $mod_opt eq "^^" ) {
$output = uc ($alt_val);
DBUG_PRINT ("MOD", "Upshift string (%s)", $output);
} elsif ( $mod_opt eq ",," ) {
$output = lc ($alt_val);
DBUG_PRINT ("MOD", "Downshift string (%s)", $output);
} elsif ( $mod_opt eq "~~" ) {
$output = $alt_val;
$output =~ s/([A-Z])|([a-z])/defined $1 ? lc($1) : uc($2)/gex;
DBUG_PRINT ("MOD", "Reverse case of each char in string (%s)", $output);
} elsif ( $mod_opt eq "^" ) {
$output = ucfirst ($alt_val);
DBUG_PRINT ("MOD", "Upshift 1st char in string (%s)", $output);
} elsif ( $mod_opt eq "," ) {
$output = lcfirst ($alt_val);
DBUG_PRINT ("MOD", "Downshift 1st char in string (%s)", $output);
} elsif ( $mod_opt eq "~" ) {
$output = ucfirst ($alt_val);
$output = lcfirst ($alt_val) if ( $alt_val eq $output );
DBUG_PRINT ("MOD", "Reverse case of 1st char in string (%s)", $output);
} else {
DBUG_PRINT ("MOD",
"The modifier (%s) didn't affect the variable's value!",
$mod_opt);
$output = $value;
}
DBUG_RETURN ( $output, $mask );
}
# ==============================================================
=item @ret[0..7] = parse_for_variables ( $value, $ignore_disable_flag, $rOpts )
This is a helper method to F<expand_variables> and B<parse_line>.
This method parses the I<$value> to see if any variables are defined in it
and returns the information about it. If there is more than one variable
present in the I<$value>, only the 1st variable/tag to evaluate is returned.
By default, a variable is the tag in the I<$value> between B<${> and B<}>, which
can be overridden with other anchor patterns. See L<Advanced::Config::Options>
for more details on this.
If you've configured the module to ignore variables, it will never find any.
Unless you also set I<$ignore_disable_flag> to a non-zero value.
Returns B<8> values. ( $left, $tag, $right, $cmt, $sub_tag, $sub_opr, $sub_val,
$otag )
All B<8> values will be I<undef> if no variables were found in I<$value>.
Otherwise it returns at least the 1st four values. Where I<$tag> is the
variable that needs to be looked up. And the caller can join things back
together as "B<$left . $look_up_value . $right>" after the variable substitution
is done and before this method is called again to locate additional variables in
the resulting new I<$value>.
The 4th value I<$cmt>, will be true/false based on if B<$left> has a comment
symbol in it! This flag only has meaning to B<parse_line>. And is terribly
misleading to other users.
Should the I<$tag> definition have one of the supported shell script variable
modifiers embedded inside it, then the I<$tag> will be parsed and the 3 B<sub_*>
return values will be calculated as well. See
L<http://wiki.bash-hackers.org/syntax/pe> for more details. Most of the
modifiers listed there are supported except for those dealing with arrays.
See I<apply_modifier> for applying these rules against the returned I<$tag>.
Other modifier rules may be added upon request.
These 3 B<sub_*> return values will always be I<undef> should the variable
left/right anchors be overridden with the same value. Or if no modifiers
are detected in the tag's name.
If you've configured the module to be case insensitive (option B<tag_case>),
then both I<$tag> and I<$sub_tag> will be shifted to lower case for case
insensitive variable lookups.
Finally there is an 8th return value, I<$otag>, that contains the original
I<$tag> value before it was edited. Needed by F<parse_line> logic.
=cut
# WARNING: If (${lvar} == ${rvar}), nested variables are not supported.
# : And neither are variable modifiers. (The sub_* return values.)
# : So evaluate tags left to right.
# : If (${lvar} != ${rvar}), nested variables are supported.
# : So evaluate inner most tags first. And then left to right.
#
# RETURNS: 8 values. ( $left, $tag, $right, $cmt, $sub_tag, $sub_opr, $sub_val, $otag )
# : The 3 sub_* vars are usually undef.
# : But when set, all 3 sub_* vars are set! And $tag != $sub_tag.
#
# NOTE 1 : If the 3 sub_* vars are populated, you'd get something like this
# : for the tag & sub_* vars.
# : tag : "abc:-Default Value" - the ${...} was removed.
# : sub_tag : "abc" - the ${...} & modifier were removed.
# : sub_opr : ":-"
# : sub_val : "Default Value"
# : So if the "tag" exists as a variable, the sub_* values are ignored.
# : But if "tag" doesn't exist as a variable, then we apply the
# : sub_* rules!
#
# NOTE 2 : If the sub_* vars undef, you'd get something like this without any
# : modifiers.
# : tag : tag - the ${...} was removed.
#
# NOTE 3 : For some alternate variable anchors, the sub_* vars will almost
# : always be undef. Since the code base won't allow you to redefine
# : these modifiers when they conflict with the variable anchors.
sub parse_for_variables
{
DBUG_ENTER_FUNC ( @_ );
my $value = shift;
my $disable_flag = shift;
my $opts = shift;
my ($left, $s1, $tag, $s2, $right, $otag);
my $cmt_flg = 0;
my ($sub_tag, $sub_opr, $sub_val, $sub_extra);
if ( $opts->{disable_variables} && (! $disable_flag) ) {
DBUG_PRINT ("INFO", "Variable substitution has been disabled.");
return DBUG_RETURN ( $left, $tag, $right, $cmt_flg,
$sub_tag, $sub_opr, $sub_val, $otag );
}
my $lvar = convert_to_regexp_string ($opts->{variable_left}, 1);
my $rvar = convert_to_regexp_string ($opts->{variable_right}, 1);
# Break up the value into it's component parts. (Non-greedy RegExpr)
if ( $value =~ m/(^.*?)(${lvar})(.*?)(${rvar})(.*$)/ ) {
($left, $s1, $tag, $s2, $right) = ($1, $2, $3, $4, $5);
$otag = $tag;
# Did a comment symbol apear before the 1st ${lvar} in the line?
my $cmt_str = convert_to_regexp_string ($opts->{comment}, 1);
$cmt_flg = 1 if ( $left =~ m/${cmt_str}/ );
DBUG_PRINT ("XXXX", "%s ===> %s <=== %s -- %d",
$left, $tag, $right, $cmt_flg);
# We know we found the 1st right hand anchor in the string's value.
# But since variables may be nested, we might not be at the correct
# left hand anchor. But at least we know they're going to balance!
# Check for nested variables ... (trim left side)
while ( $tag =~ m/(^.*)${lvar}(.*?$)/ ) {
my ($l, $t) = ($1, $2);
$left .= $s1 . $l;
$tag = $t;
}
# Strip off leading spaces from the tag's name.
# No tag may have leading spaces in it.
# Defering the stripping of trailing spaces until later on purpose!
$tag =~ s/^\s+//;
# -----------------------------------------------------------
# We have a variable! Now check if there are modifiers
# in it that we are supporting ...
# See: http://wiki.bash-hackers.org/syntax/pe
# -----------------------------------------------------------
# The variable modifier tags. Needed to avoid using the wrong rule.
# A variable name can use anything except for what's in this list!
my $not = "[^-:?+#%/\^,~]";
if ( $lvar eq $rvar ) {
; # No modifiers are supported if the left/right anchors are the same!
# Since there are too many modifier/anchor pairs that no longer
# work. Behaving more like a Windows *.bat file now.
} elsif ( $opts->{disable_variable_modifiers} ) {
; # Explicitly told not to use this feature.
# Rule: :-, :=, :+, -, =, or +
} elsif ( $tag =~ m/(^${not}+)(:?[-=+])(.+)$/) {
($sub_tag, $sub_opr, $sub_val) = ($1, $2, $3);
# Rule: :? or ?
} elsif ( $tag =~ m/(^${not}+)(:?[?])(.*)$/) {
( run in 2.062 seconds using v1.01-cache-2.11-cpan-d8267643d1d )