Specify the set of benches to run. The following grammar defines the
I<directives>:
- I<directives> ::= I<directives> | I<directives> -- Alternation
- | I<directives> & I<directives> -- Concatenation
- | [ I<directives> ] -- Optional
- | ( I<directives> ) -- Parentheses
- | I<directive>
+ directives ::=
+ directives | directives -- Alternation
+ | directives & directives -- Concatenation
+ | [ directives> ] -- Optional
+ | ( directives> ) -- Parentheses
+ | #d NAME[=VALUE] -- %code { #define NAME [VALUE] }
+ | %d NAME[=VALUE] -- %define NAME ["VALUE"]
+ | %s skeleton -- %skeleton "skeleton"
+ | directive
Parentheses only group to override precedence. For instance:
{
my ($base, $max, @directive) = @_;
my $directives = directives ($base, @directive);
- my $variant = grep { $_ eq '%define variant' } @directive;
-
+ my $variant = grep { /%define "?variant"?/ } @directive;
+ my $lex_symbol = grep { /%define "?lex_symbol"?/ } @directive;
my $out = new IO::File ">$base.y"
or die;
print $out <<EOF;
%language "C++"
%defines
+%locations
$directives
%code requires // *.h
#include <iostream>
#include <sstream>
-// Prototype of the yylex function providing subsequent tokens.
-static yy::parser::token_type yylex(yy::parser::semantic_type* yylval);
-
#define STAGE_MAX ($max * 10) // max = $max
+
+#define USE_LEX_SYMBOL $lex_symbol
#define USE_VARIANTS $variant
-#if USE_VARIANTS
-# define IF_VARIANTS(True, False) True
+
+ // Prototype of the yylex function providing subsequent tokens.
+ static
+#if USE_LEX_SYMBOL
+ yy::parser::symbol_type yylex();
#else
-# define IF_VARIANTS(True, False) False
+ yy::parser::token_type yylex(yy::parser::semantic_type* yylval,
+ yy::parser::location_type* yylloc);
#endif
+
+ // Conversion to string.
+ template <typename T>
+ inline
+ std::string
+ string_cast (const T& t)
+ {
+ std::ostringstream o;
+ o << t;
+ return o.str ();
+ }
}
+
+%token END_OF_FILE 0
EOF
if ($variant)
%token <int> NUMBER
%printer { std::cerr << "Number: " << $$; } <int>
%printer { std::cerr << "Text: " << $$; } <std::string>
-%token END_OF_FILE 0
%type <std::string> text result
%%
text:
/* nothing */ { /* This will generate an empty string */ }
-| text TEXT { std::swap($$,$1); $$.append($2); }
-| text NUMBER {
- std::swap($$,$1);
- std::ostringstream ss;
- ss << ' ' << $2;
- $$.append(ss.str());
- }
+| text TEXT { std::swap ($$, $2); }
+| text NUMBER { $$ = string_cast($2); }
;
EOF
}
%token <ival> NUMBER
%printer { std::cerr << "Number: " << $$; } <ival>
%printer { std::cerr << "Text: " << *$$; } <sval>
-%token END_OF_FILE 0
%type <sval> text result
%%
text:
/* nothing */ { $$ = new std::string; }
-| text TEXT { $$->append(*$2); delete $2; }
-| text NUMBER {
- std::ostringstream ss;
- ss << ' ' << $2;
- $$->append(ss.str());
- }
+| text TEXT { delete $1; $$ = $2; }
+| text NUMBER { delete $1; $$ = new std::string (string_cast ($2)); }
;
EOF
}
print $out <<'EOF';
%%
+#
+
static
-yy::parser::token_type
-yylex(yy::parser::semantic_type* yylval)
+#if USE_LEX_SYMBOL
+yy::parser::symbol_type yylex()
+#else
+yy::parser::token_type yylex(yy::parser::semantic_type* yylval,
+ yy::parser::location_type* yylloc)
+#endif
{
+ typedef yy::parser::location_type location_type;
+ typedef yy::parser::token token;
static int stage = -1;
++stage;
if (stage == STAGE_MAX)
- return yy::parser::token::END_OF_FILE;
+ {
+#if USE_LEX_SYMBOL
+ return yy::parser::make_END_OF_FILE (yy::location());
+#else
+ *yylloc = location_type ();
+ return token::END_OF_FILE;
+#endif
+ }
else if (stage % 2)
{
- IF_VARIANTS(yylval->build<int>(), yylval->ival) = stage;
- return yy::parser::token::NUMBER;
+#if USE_LEX_SYMBOL
+ return yy::parser::make_NUMBER (stage, yy::location());
+#else
+# if defined ONE_STAGE_BUILD
+ yylval->build(stage);
+# elif USE_VARIANTS
+ yylval->build<int>() = stage;
+# else
+ yylval->ival = stage;
+# endif
+ *yylloc = location_type ();
+ return token::NUMBER;
+#endif
}
else
{
- IF_VARIANTS(yylval->build<std::string>() =, yylval->sval = new) std::string("A string.");
- return yy::parser::token::TEXT;
+#if USE_LEX_SYMBOL
+ return yy::parser::make_TEXT ("A string.", yy::location());
+#else
+# if defined ONE_STAGE_BUILD
+ yylval->build(std::string("A string."));
+# elif USE_VARIANTS
+ yylval->build<std::string>() = std::string("A string.");
+# else
+ yylval->sval = new std::string("A string.");
+# endif
+ *yylloc = location_type ();
+ return token::TEXT;
+#endif
}
abort();
}
sub generate_grammar ($$@)
{
my ($name, $base, @directive) = @_;
- verbose 2, "Generating $base.y\n";
+ verbose 3, "Generating $base.y\n";
my %generator =
(
"calc" => \&generate_grammar_calc,
sub run ($)
{
my ($command) = @_;
- verbose 2, "$command\n";
+ verbose 3, "$command\n";
system ("$command") == 0
or die "$command failed";
}
# shows only wallclock and the two children times. 'auto' (the
# default) will act as 'all' unless the children times are both
# zero, in which case it acts as 'noc'. 'none' prevents output.
- verbose 2, "Running the benches for $grammar\n";
+ verbose 3, "Running the benches for $grammar\n";
my $res = timethese ($iterations, \%bench, 'nop');
# Output the speed result.
sub bench_push_parser ()
{
bench ('calc',
- (
- '[', '%define api.pure', ']',
- '&',
- '[', '%define api.push_pull "both"', ']'
+ qw(
+ [ %d api.pure ]
+ &
+ [ %d api.push_pull=both ]
));
}
=item C<bench_variant_parser ()>
-Bench the C++ lalr1.cc parser using Boost.Variants or %union.
+Bench the C++ lalr1.cc parser using variants or %union.
=cut
sub bench_variant_parser ()
{
- bench ('variant',
- ('%skeleton "lalr1.cc"',
- '&',
- '[', '%debug', ']',
- '&',
- '[', '%define variant', ']',
- '&',
- '[', "%code {\n#define VARIANT_DESTROY\n}", ']'
- ));
+ bench ('list',
+ qw(
+ [
+ %d variant
+ &
+ [ #d ONE_STAGE_BUILD | %d lex_symbol ]
+ ]
+ )
+ );
}
######################################################################
sub bench_fusion_parser ()
{
bench ('list',
- ('%skeleton "lalr1-split.cc"',
- '|',
- '%skeleton "lalr1.cc"'));
+ qw(
+ %s lalr1-split.cc
+ |
+ %s lalr1.cc
+ )
+ );
}
############################################################################
######################################################################
+# The end of the directives to parse.
+my $eod = "end of directives";
# The list of tokens parsed by the following functions.
my @token;
+# eat ($EXPECTED)
+# ---------------
+# Check that the current token is $EXPECTED, and move to the next.
+sub eat ($)
+{
+ my ($expected) = @_;
+ die "expected $expected, unexpected: $token[0] (@token)\n"
+ unless $token[0] eq $expected;
+ shift @token;
+}
+
# Parse directive specifications:
# expr: term (| term)*
# term: fact (& fact)*
# fact: ( expr ) | [ expr ] | dirs
+# dirs: %s SKELETON | #d NAME[=VALUE] | %d NAME[=VALUE] | directive
sub parse (@)
{
- @token = @_;
- verbose 2, "Parsing: @token\n";
- return parse_expr ();
+ @token = (@_, $eod);
+ verbose 3, "Parsing: @token\n";
+ my @res = parse_expr ();
+ eat ($eod);
+ return @res;
}
sub parse_expr ()
{
my @res = parse_term ();
- while (defined $token[0] && $token[0] eq '|')
+ while ($token[0] eq '|')
{
- shift @token;
+ eat ('|');
# Alternation.
push @res, parse_term ();
}
sub parse_term ()
{
my @res = parse_fact ();
- while (defined $token[0] && $token[0] eq '&')
+ while ($token[0] eq '&')
{
- shift @token;
+ eat ('&');
# Cartesian product.
my @lhs = @res;
@res = ();
{
for my $lhs (@lhs)
{
- push @res, "$lhs\n$rhs";
+ push @res, $lhs . ($lhs && $rhs ? "\n" : "") . $rhs;
}
}
}
if ($token[0] eq '(')
{
- shift @token;
+ eat ('(');
@res = parse_expr ();
- die "unexpected $token[0], expected )"
- unless $token[0] eq ')';
- shift @token;
+ eat (')');
}
elsif ($token[0] eq '[')
{
- shift @token;
+ eat ('[');
@res = (parse_expr (), '');
- die "unexpected $token[0], expected ]"
- unless $token[0] eq ']';
+ eat (']');
+ }
+ else
+ {
+ @res = parse_dirs ();
+ }
+ return @res;
+}
+
+sub parse_dirs ()
+{
+ my @res;
+ die "unexpected end of expression"
+ unless defined $token[0];
+
+ if ($token[0] eq '#d')
+ {
+ eat ('#d');
+ $token[0] =~ s/(.*?)=(.*)/$1 $2/;
+ @res = ("%code {\n#define $token[0]\n}");
+ shift @token;
+ }
+ elsif ($token[0] eq '%d')
+ {
+ shift @token;
+ $token[0] =~ s/(.*?)=(.*)/$1 "$2"/;
+ @res = ("%define $token[0]");
+ shift @token;
+ }
+ elsif ($token[0] eq '%s')
+ {
+ shift @token;
+ @res = ("%skeleton \"$token[0]\"");
shift @token;
}
else
@res = $token[0];
shift @token;
}
+
return @res;
}
Getopt::Long::Configure ("bundling", "pass_through");
GetOptions (%option)
or exit 1;
-
- # Support -b: predefined benches.
- my %bench =
- (
- "fusion" => \&bench_fusion_parser,
- "push" => \&bench_push_parser,
- "variant" => \&bench_variant_parser,
- );
-
- if (defined $bench)
- {
- die "invalid argument for --bench: $bench"
- unless defined $bench{$bench};
- &{$bench{$bench}}();
- exit 0;
- }
}
######################################################################
getopt;
+
+# Create the directory we work in.
+mkdir "benches" or die "cannot create benches"
+ unless -d "benches";
+my $count = 1;
+++$count
+ while -d "benches/$count";
+my $dir = "benches/$count";
+mkdir $dir
+ or die "cannot create $dir";
+chdir $dir
+ or die "cannot chdir $dir";
+
+# The following message is tailored to please Emacs' compilation-mode.
+verbose 1, "Entering directory `$dir'\n";
verbose 1, "Using bison=$bison.\n";
-verbose 1, "Using cc=$cc.\n";
-verbose 1, "Using cxx=$cxx.\n";
-verbose 1, "Using cflags=$cflags.\n";
+verbose 2, "Using cc=$cc.\n";
+verbose 2, "Using cxx=$cxx.\n";
+verbose 2, "Using cflags=$cflags.\n";
verbose 2, "Grammar: $grammar\n";
-# Launch the bench marking.
-bench ($grammar, @ARGV);
+# Support -b: predefined benches.
+my %bench =
+ (
+ "fusion" => \&bench_fusion_parser,
+ "push" => \&bench_push_parser,
+ "variant" => \&bench_variant_parser,
+ );
+
+if (defined $bench)
+{
+ die "invalid argument for --bench: $bench"
+ unless defined $bench{$bench};
+ &{$bench{$bench}}();
+ exit 0;
+}
+else
+{
+ # Launch the bench marking.
+ bench ($grammar, @ARGV);
+}
### Setup "GNU" style for perl-mode and cperl-mode.
## Local Variables:
projects / bison.git / blobdiff