User manual (re2c)


A re2c program consists of normal code intermixed with re2c blocks and directives. Each re2c block may contain definitions, configurations and rules. Definitions are of the form name = regexp; where name is an identifier that consists of letters, digits and underscores, and regexp is a regular expression. Regular expressions may contain other definitions, but recursion is not allowed and each name should be defined before used. Configurations are of the form re2c:config = value; where config is the configuration descriptor and value can be a number, a string or a special word. Rules consist of a regular expression followed by a semantic action (a block of code enclosed in curly braces { and }, or a raw one line of code preceded with := and ended with a newline that is not followed by a whitespace). If the input matches the regular expression, the associated semantic action is executed. If multiple rules match, the longest match takes precedence. If multiple rules match the same string, the earlier rule takes precedence. There are two special rules: default rule * and EOF rule $. Default rule should always be defined, it has the lowest priority regardless of its place and matches any code unit (not necessarily a valid character, see encoding support). EOF rule matches the end of input, it should be defined if the corresponding EOF handling method is used. If start conditions are used, rules have more complex syntax. All rules of a single block are compiled into a deterministic finite-state automaton (DFA) and encoded in the form of a program in the target language. The generated code interfaces with the outer program by the means of a few user-defined primitives (see the program interface section). Reusable blocks allow sharing rules, definitions and configurations between different blocks.

Here is an example program that shows various aspects of re2c syntax:

// re2c $INPUT -o $OUTPUT -i
#include <assert.h>                 //
                                    // C/C++ code
int lex(const char *YYCURSOR)       //
{
    /*!re2c                         // start of re2c block
    re2c:define:YYCTYPE = char;     // configuration
    re2c:yyfill:enable = 0;         // configuration
    re2c:flags:case-ranges = 1;     // configuration
                                    //
    ident = [a-zA-Z_][a-zA-Z_0-9]*; // named definition
                                    //
    ident { return 0; }             // normal rule
    *     { return 1; }             // default rule
    */
}                                   //
                                    //
int main()                          //
{                                   // C/C++ code
    assert(lex("_Zer0") == 0);      //
    return 0;                       //
}                                   //

This is the generated output:

/* Generated by re2c */
// re2c $INPUT -o $OUTPUT -i
#include <assert.h>                 //
                                    // C/C++ code
int lex(const char *YYCURSOR)       //
{
    
{
	char yych;
	yych = *YYCURSOR;
	switch (yych) {
	case 'A' ... 'Z':
	case '_':
	case 'a' ... 'z': goto yy4;
	default: goto yy2;
	}
yy2:
	++YYCURSOR;
	{ return 1; }
yy4:
	yych = *++YYCURSOR;
	switch (yych) {
	case '0' ... '9':
	case 'A' ... 'Z':
	case '_':
	case 'a' ... 'z': goto yy4;
	default: goto yy6;
	}
yy6:
	{ return 0; }
}

}                                   //
                                    //
int main()                          //
{                                   // C/C++ code
    assert(lex("_Zer0") == 0);      //
    return 0;                       //
}                                   //

Command-line options

-? -h --help

Show help message.

-1 --single-pass

Deprecated. Does nothing (single pass is the default now).

-8 --utf-8

Generate a lexer that reads input in UTF-8 encoding. re2c assumes that character range is 0 – 0x10FFFF and character size is 1 byte.

-b --bit-vectors

Optimize conditional jumps using bit masks. Implies -s.

-c --conditions --start-conditions

Enable support of Flex-like “conditions”: multiple interrelated lexers within one block. Option --start-conditions is a legacy alias; use --conditions instead.

--case-insensitive

Treat single-quoted and double-quoted strings as case-insensitive.

--case-inverted

Invert the meaning of single-quoted and double-quoted strings: treat single-quoted strings as case-sensitive and double-quoted strings as case-insensitive.

--case-ranges

Collapse consecutive cases in a switch statements into a range of the form case low ... high:. This syntax is an extension of the C/C++ language, supported by compilers like GCC, Clang and Tcc. The main advantage over using single cases is smaller generated C code and faster generation time, although for some compilers like Tcc it also results in smaller binary size. This option doesn’t work for the Go backend.

-e --ecb

Generate a lexer that reads input in EBCDIC encoding. re2c assumes that character range is 0 – 0xFF an character size is 1 byte.

--empty-class <match-empty | match-none | error>

Define the way re2c treats empty character classes. With match-empty (the default) empty class matches empty input (which is illogical, but backwards-compatible). With``match-none`` empty class always fails to match. With error empty class raises a compilation error.

--encoding-policy <fail | substitute | ignore>

Define the way re2c treats Unicode surrogates. With fail re2c aborts with an error when a surrogate is encountered. With substitute re2c silently replaces surrogates with the error code point 0xFFFD. With ignore (the default) re2c treats surrogates as normal code points. The Unicode standard says that standalone surrogates are invalid, but real-world libraries and programs behave in different ways.

-f --storable-state

Generate a lexer which can store its inner state. This is useful in push-model lexers which are stopped by an outer program when there is not enough input, and then resumed when more input becomes available. In this mode users should additionally define YYGETSTATE() and YYSETSTATE(state) macros and variables yych, yyaccept and state as part of the lexer state.

-F --flex-syntax

Partial support for Flex syntax: in this mode named definitions don’t need the equal sign and the terminating semicolon, and when used they must be surrounded by curly braces. Names without curly braces are treated as double-quoted strings.

-g --computed-gotos

Optimize conditional jumps using non-standard “computed goto” extension (which must be supported by the compiler). re2c generates jump tables only in complex cases with a lot of conditional branches. Complexity threshold can be configured with cgoto:threshold configuration. This option implies -b. This option doesn’t work for the Go backend.

-I PATH

Add PATH to the list of locations which are used when searching for include files. This option is useful in combination with /*!include:re2c ... */ directive. Re2c looks for FILE in the directory of including file and in the list of include paths specified by -I option.

-i --no-debug-info

Do not output #line information. This is useful when the generated code is tracked by some version control system or IDE.

--input <default | custom>

Specify the API used by the generated code to interface with used-defined code. Option default is the C API based on pointer arithmetic (it is the default for the C backend). Option custom is the generic API (it is the default for the Go backend).

--input-encoding <ascii | utf8>

Specify the way re2c parses regular expressions. With ascii (the default) re2c handles input as ASCII-encoded: any sequence of code units is a sequence of standalone 1-byte characters. With utf8 re2c handles input as UTF8-encoded and recognizes multibyte characters.

--lang <c | go>

Specify the output language. Supported languages are C and Go (the default is C).

--location-format <gnu | msvc>

Specify location format in messages. With gnu locations are printed as ‘filename:line:column: …’. With msvc locations are printed as ‘filename(line,column) …’. Default is gnu.

--no-generation-date

Suppress date output in the generated file.

--no-version

Suppress version output in the generated file.

-o OUTPUT --output=OUTPUT

Specify the OUTPUT file.

-P --posix-captures

Enable submatch extraction with POSIX-style capturing groups.

-r --reusable

Allows reuse of re2c rules with /*!rules:re2c */ and /*!use:re2c */ blocks. Exactly one rules-block must be present. The rules are saved and used by every use-block that follows, which may add its own rules and configurations.

-S --skeleton

Ignore user-defined interface code and generate a self-contained “skeleton” program. Additionally, generate input files with strings derived from the regular grammar and compressed match results that are used to verify “skeleton” behavior on all inputs. This option is useful for finding bugs in optimizations and code generation. This option doesn’t work for the Go backend.

-s --nested-ifs

Use nested if statements instead of switch statements in conditional jumps. This usually results in more efficient code with non-optimizing compilers.

-T --tags

Enable submatch extraction with tags.

-t HEADER --type-header=HEADER

Generate a HEADER file that contains enum with condition names. Requires -c option.

-u --unicode

Generate a lexer that reads UTF32-encoded input. Re2c assumes that character range is 0 – 0x10FFFF and character size is 4 bytes. This option implies -s.

-V --vernum

Show version information in MMmmpp format (major, minor, patch).

--verbose

Output a short message in case of success.

-v --version

Show version information.

-w --wide-chars

Generate a lexer that reads UCS2-encoded input. Re2c assumes that character range is 0 – 0xFFFF and character size is 2 bytes. This option implies -s.

-x --utf-16

Generate a lexer that reads UTF16-encoded input. Re2c assumes that character range is 0 – 0x10FFFF and character size is 2 bytes. This option implies -s.


Debug options:

-D --emit-dot

Instead of normal output generate lexer graph in .dot format. The output can be converted to an image with the help of Graphviz (e.g. something like dot -Tpng -odfa.png dfa.dot).

-d --debug-output

Emit YYDEBUG in the generated code. YYDEBUG should be defined by the user in the form of a void function with two parameters: state (lexer state or -1) and symbol (current input symbol of type YYCTYPE).

--dump-adfa

Debug option: output DFA after tunneling (in .dot format).

--dump-cfg

Debug option: output control flow graph of tag variables (in .dot format).

--dump-closure-stats

Debug option: output statistics on the number of states in closure.

--dump-dfa-det

Debug option: output DFA immediately after determinization (in .dot format).

--dump-dfa-min

Debug option: output DFA after minimization (in .dot format).

--dump-dfa-tagopt

Debug option: output DFA after tag optimizations (in .dot format).

--dump-dfa-tree

Debug option: output DFA under construction with states represented as tag history trees (in .dot format).

--dump-dfa-raw

Debug option: output DFA under construction with expanded state-sets (in .dot format).

--dump-interf

Debug option: output interference table produced by liveness analysis of tag variables.

--dump-nfa

Debug option: output NFA (in .dot format).


Internal options:

--dfa-minimization <moore | table>

Internal option: DFA minimization algorithm used by re2c. The moore option is the Moore algorithm (it is the default). The table option is the “table filling” algorithm. Both algorithms should produce the same DFA up to states relabeling; table filling is simpler and much slower and serves as a reference implementation.

--eager-skip

Internal option: make the generated lexer advance the input position eagerly – immediately after reading the input symbol. This changes the default behavior when the input position is advanced lazily – after transition to the next state. This option is implied by --no-lookahead.

--no-lookahead

Internal option: use TDFA(0) instead of TDFA(1). This option has effect only with --tags or --posix-captures options.

--no-optimize-tags

Internal optionL: suppress optimization of tag variables (useful for debugging).

--posix-closure <gor1 | gtop>

Internal option: specify shortest-path algorithm used for the construction of epsilon-closure with POSIX disambiguation semantics: gor1 (the default) stands for Goldberg-Radzik algorithm, and gtop stands for “global topological order” algorithm.

--posix-prectable <complex | naive>

Internal option: specify the algorithm used to compute POSIX precedence table. The complex algorithm computes precedence table in one traversal of tag history tree and has quadratic complexity in the number of TNFA states; it is the default. The naive algorithm has worst-case cubic complexity in the number of TNFA states, but it is much simpler than complex and may be slightly faster in non-pathological cases.

--stadfa

Internal option: use staDFA algorithm for submatch extraction. The main difference with TDFA is that tag operations in staDFA are placed in states, not on transitions.


Warnings:

-W

Turn on all warnings.

-Werror

Turn warnings into errors. Note that this option alone doesn’t turn on any warnings; it only affects those warnings that have been turned on so far or will be turned on later.

-W<warning>

Turn on warning.

-Wno-<warning>

Turn off warning.

-Werror-<warning>

Turn on warning and treat it as an error (this implies -W<warning>).

-Wno-error-<warning>

Don’t treat this particular warning as an error. This doesn’t turn off the warning itself.


Individual warnings (see the detailed descriptions):

-Wcondition-order

Warn if the generated program makes implicit assumptions about condition numbering. One should use either the -t, --type-header option or the /*!types:re2c*/ directive to generate a mapping of condition names to numbers and then use the autogenerated condition names.

-Wempty-character-class

Warn if a regular expression contains an empty character class. Trying to match an empty character class makes no sense: it should always fail. However, for backwards compatibility reasons re2c allows empty character classes and treats them as empty strings. Use the --empty-class option to change the default behavior.

-Wmatch-empty-string

Warn if a rule is nullable (matches an empty string). If the lexer runs in a loop and the empty match is unintentional, the lexer may unexpectedly hang in an infinite loop.

-Wswapped-range

Warn if the lower bound of a range is greater than its upper bound. The default behavior is to silently swap the range bounds.

-Wundefined-control-flow

Warn if some input strings cause undefined control flow in the lexer (the faulty patterns are reported). This is the most dangerous and most common mistake. It can be easily fixed by adding the default rule * which has the lowest priority, matches any code unit, and consumes exactly one code unit.

-Wunreachable-rules

Warn about rules that are shadowed by other rules and will never match.

-Wuseless-escape

Warn if a symbol is escaped when it shouldn’t be. By default, re2c silently ignores such escapes, but this may as well indicate a typo or an error in the escape sequence.

-Wnondeterministic-tags

Warn if a tag has n-th degree of nondeterminism, where n is greater than 1.

-Wsentinel-in-midrule

Warn if the sentinel symbol occurs in the middle of a rule — this may cause reads past the end of buffer, crashes or memory corruption in the generated lexer. This warning is only applicable if the sentinel method of checking for the end of input is used. It is set to an error if re2c:sentinel configuration is used.

Program interface

Re2c has a flexible interface that gives the user both the freedom and the responsibility to define how the generated code interacts with the outer program. There are two major options:

  • Pointer API. It is also called “default API”, since it was historically the first, and for a long time the only one. This is a more restricted API based on C pointer arithmetics. It consists of pointer-like primitives YYCURSOR, YYMARKER, YYCTXMARKER and YYLIMIT, which are normally defined as pointers of type YYCTYPE*. Pointer API is enabled by default for the C backend, and it cannot be used with other backends that do not have pointer arithmetics.


  • Generic API. This is a less restricted API that does not assume pointer semantics. It consists of primitives YYPEEK, YYSKIP, YYBACKUP, YYBACKUPCTX, YYSTAGP, YYSTAGN, YYMTAGP, YYMTAGN, YYRESTORE, YYRESTORECTX, YYRESTORETAG, YYSHIFT, YYSHIFTSTAG, YYSHIFTMTAG and YYLESSTHAN. For the C backend generic API is enabled with --input custom option or re2c:flags:input = custom; configuration; for the Go backend it is enabled by default. Generic API was added in version 0.14. It is intentionally designed to give the user as much freedom as possible in redefining the input model and the semantics of different actions performed by the generated code. As an example, one can override YYPEEK to check for the end of input before reading the input character, or do some logging, etc.

Generic API has two styles:

  • Function-like. This style is enabled with re2c:api:style = functions; configuration, and it is the default for C backend. In this style API primitives should be defined as functions or macros with parentheses, accepting the necessary arguments. For example, in C the default pointer API can be defined in function-like style generic API as follows:

    #define  YYPEEK()                 *YYCURSOR
    #define  YYSKIP()                 ++YYCURSOR
    #define  YYBACKUP()               YYMARKER = YYCURSOR
    #define  YYBACKUPCTX()            YYCTXMARKER = YYCURSOR
    #define  YYRESTORE()              YYCURSOR = YYMARKER
    #define  YYRESTORECTX()           YYCURSOR = YYCTXMARKER
    #define  YYRESTORETAG(tag)        YYCURSOR = tag
    #define  YYLESSTHAN(len)          YYLIMIT - YYCURSOR < len
    #define  YYSTAGP(tag)             tag = YYCURSOR
    #define  YYSTAGN(tag)             tag = NULL
    #define  YYSHIFT(shift)           YYCURSOR += shift
    #define  YYSHIFTSTAG(tag, shift)  tag += shift
    

  • Free-form. This style is enabled with re2c:api:style = free-form; configuration, and it is the default for Go backend. In this style API primitives can be defined as free-form pieces of code, and instead of arguments they have interpolated variables of the form @@{name}, or optionally just @@ if there is only one argument. The @@ text is called “sigil”. It can be redefined to any other text with re2c:api:sigil configuration. For example, the default pointer API can be defined in free-form style generic API as follows:

    re2c:define:YYPEEK       = "*YYCURSOR";
    re2c:define:YYSKIP       = "++YYCURSOR";
    re2c:define:YYBACKUP     = "YYMARKER = YYCURSOR";
    re2c:define:YYBACKUPCTX  = "YYCTXMARKER = YYCURSOR";
    re2c:define:YYRESTORE    = "YYCURSOR = YYMARKER";
    re2c:define:YYRESTORECTX = "YYCURSOR = YYCTXMARKER";
    re2c:define:YYRESTORETAG = "YYCURSOR = ${tag}";
    re2c:define:YYLESSTHAN   = "YYLIMIT - YYCURSOR < @@{len}";
    re2c:define:YYSTAGP      = "@@{tag} = YYCURSOR";
    re2c:define:YYSTAGN      = "@@{tag} = NULL";
    re2c:define:YYSHIFT      = "YYCURSOR += @@{shift}";
    re2c:define:YYSHIFTSTAG  = "@@{tag} += @@{shift}";
    

API primitives

Here is a list of API primitives that may be used by the generated code in order to interface with the outer program. Which primitives are needed depends on multiple factors, including the complexity of regular expressions, input representation, buffering, the use of various features and so on. All the necessary primitives should be defined by the user in the form of macros, functions, variables, free-form pieces of code or any other suitable form. Re2c does not (and cannot) check the definitions, so if anything is missing or defined incorrectly the generated code will not compile.

YYCTYPE

The type of the input characters (code units). For ASCII, EBCDIC and UTF-8 encodings it should be 1-byte unsigned integer. For UTF-16 or UCS-2 it should be 2-byte unsigned integer. For UTF-32 it should be 4-byte unsigned integer.

YYCURSOR

A pointer-like l-value that stores the current input position (usually a pointer of type YYCTYPE*). Initially YYCURSOR should point to the first input character. It is advanced by the generated code. When a rule matches, YYCURSOR points to the one after the last matched character. It is used only in the default C API.

YYLIMIT

A pointer-like r-value that stores the end of input position (usually a pointer of type YYCTYPE*). Initially YYLIMIT should point to the one after the last available input character. It is not changed by the generated code. Lexer compares YYCURSOR to YYLIMIT in order to determine if there is enough input characters left. YYLIMIT is used only in the default C API.

YYMARKER

A pointer-like l-value (usually a pointer of type YYCTYPE*) that stores the position of the latest matched rule. It is used to restores YYCURSOR position if the longer match fails and lexer needs to rollback. Initialization is not needed. YYMARKER is used only in the default C API.

YYCTXMARKER

A pointer-like l-value that stores the position of the trailing context (usually a pointer of type YYCTYPE*). No initialization is needed. It is used only in the default C API, and only with the lookahead operator /.

YYFILL

API primitive with one argument len. The meaning of YYFILL is to provide at least len more input characters or fail. If EOF rule is used, YYFILL should always return to the calling function; the return value should be zero on success and non-zero on failure. If EOF rule is not used, YYFILL return value is ignored and it should not return on failure. Maximal value of len is YYMAXFILL, which can be generated with /*!max:re2c*/ directive. The definition of YYFILL can be either function-like or free-form depending on the API style (see re2c:api:style and re2c:define:YYFILL:naked).

YYMAXFILL

An integral constant equal to the maximal value of YYFILL argument. It can be generated with /*!max:re2c*/ directive.

YYLESSTHAN

A generic API primitive with one argument len. It should be defined as an r-value of boolean type that equals true if and only if there is less than len input characters left. The definition can be either function-like or free-form depending on the API style (see re2c:api:style).

YYPEEK

A generic API primitive with no arguments. It should be defined as an r-value of type YYCTYPE that is equal to the character at the current input position. The definition can be either function-like or free-form depending on the API style (see re2c:api:style).

YYSKIP

A generic API primitive with no arguments. The meaning of YYSKIP is to advance the current input position by one character. The definition can be either function-like or free-form depending on the API style (see re2c:api:style).

YYBACKUP

A generic API primitive with no arguments. The meaning of YYBACKUP is to save the current input position, which is later restored with YYRESTORE. The definition should be either function-like or free-form depending on the API style (see re2c:api:style).

YYRESTORE

A generic API primitive with no arguments. The meaning of YYRESTORE is to restore the current input position to the value saved by YYBACKUP. The definition should be either function-like or free-form depending on the API style (see re2c:api:style).

YYBACKUPCTX

A generic API primitive with zero arguments. The meaning of YYBACKUPCTX is to save the current input position as the position of the trailing context, which is later restored by YYRESTORECTX. The definition should be either function-like or free-form depending on the API style (see re2c:api:style).

YYRESTORECTX

A generic API primitive with no arguments. The meaning of YYRESTORECTX is to restore the trailing context position saved with YYBACKUPCTX. The definition should be either function-like or free-form depending on the API style (see re2c:api:style).

YYRESTORETAG

A generic API primitive with one argument tag. The meaning of YYRESTORETAG is to restore the trailing context position to the value of tag. The definition should be either function-like or free-form depending on the API style (see re2c:api:style).

YYSTAGP

A generic API primitive with one argument tag. The meaning of YYSTAGP is to set tag value to the current input position. The definition should be either function-like or free-form depending on the API style (see re2c:api:style).

YYSTAGN

A generic API primitive with one argument tag. The meaning of YYSTAGP is to set tag value to null (or some default value). The definition should be either function-like or free-form depending on the API style (see re2c:api:style).

YYMTAGP

A generic API primitive with one argument tag. The meaning of YYMTAGP is to append the current position to the history of tag. The definition should be either function-like or free-form depending on the API style (see re2c:api:style).

YYMTAGN

A generic API primitive with one argument tag. The meaning of YYMTAGN is to append null (or some other default) value to the history of tag. The definition can be either function-like or free-form depending on the API style (see re2c:api:style).

YYSHIFT

A generic API primitive with one argument shift. The meaning of YYSHIFT is to shift the current input position by shift characters (the shift value may be negative). The definition can be either function-like or free-form depending on the API style (see re2c:api:style).

YYSHIFTSTAG

A generic API primitive with two arguments, tag and shift. The meaning of YYSHIFTSTAG is to shift tag by shift characters (the shift value may be negative). The definition can be either function-like or free-form depending on the API style (see re2c:api:style).

YYSHIFTMTAG

A generic API primitive with two arguments, tag and shift. The meaning of YYSHIFTMTAG is to shift the latest value in the history of tag by shift characters (the shift value may be negative). The definition should be either function-like or free-form depending on the API style (see re2c:api:style).

YYMAXNMATCH

An integral constant equal to the maximal number of POSIX capturing groups in a rule. It is generated with /*!maxnmatch:re2c*/ directive.

YYCONDTYPE

The type of the condition enum. It should be generated either with /*!types:re2c*/ directive or -t --type-header option.

YYGETCONDITION

An API primitive with zero arguments. It should be defined as an r-value of type YYCONDTYPE that is equal to the current condition identifier. The definition can be either function-like or free-form depending on the API style (see re2c:api:style and re2c:define:YYGETCONDITION:naked).

YYSETCONDITION

An API primitive with one argument cond. The meaning of YYSETCONDITION is to set the current condition identifier to cond. The definition should be either function-like or free-form depending on the API style (see re2c:api:style and re2c:define:YYSETCONDITION@cond).

YYGETSTATE

An API primitive with zero arguments. It should be defined as an r-value of integer type that is equal to the current lexer state. Should be initialized to -1. The definition can be either function-like or free-form depending on the API style (see re2c:api:style and re2c:define:YYGETSTATE:naked).

YYSETSTATE

An API primitive with one argument state. The meaning of YYSETSTATE is to set the current lexer state to state. The definition should be either function-like or free-form depending on the API style (see re2c:api:style and re2c:define:YYSETSTATE@state).

YYDEBUG

A debug API primitive with two arguments. It can be used to debug the generated code (with -d --debug-output option). YYDEBUG should return no value and accept two arguments: state (either a DFA state index or -1) and symbol (the current input symbol).

yych

An l-value of type YYCTYPE that stores the current input character. User definition is necessary only with -f --storable-state option.

yyaccept

An l-value of unsigned integral type that stores the number of the latest matched rule. User definition is necessary only with -f --storable-state option.

yynmatch

An l-value of unsigned integral type that stores the number of POSIX capturing groups in the matched rule. Used only with -P --posix-captures option.

yypmatch

An array of l-values that are used to hold the tag values corresponding to the capturing parentheses in the matching rule. Array length must be at least yynmatch * 2 (usually YYMAXNMATCH * 2 is a good choice). Used only with -P --posix-captures option.

Directives

Below is the list of all directives provided by re2c (in no particular order). More information on each directive can be found in the related sections.

/*!re2c ... */

A standard re2c block.

%{ ... %}

A standard re2c block in -F --flex-support mode.

/*!rules:re2c ... */

A reusable re2c block (requires -r --reuse option).

/*!use:re2c ... */

A block that reuses previous rules-block specified with /*!rules:re2c ... */ (requires -r --reuse option).

/*!ignore:re2c ... */

A block which contents are ignored and cut off from the output file.

/*!max:re2c*/

This directive is substituted with the macro-definition of YYMAXFILL.

/*!maxnmatch:re2c*/

This directive is substituted with the macro-definition of YYMAXNMATCH (requires -P --posix-captures option).

/*!getstate:re2c*/

This directive is substituted with conditional dispatch on lexer state (requires -f --storable-state option).

/*!types:re2c ... */

This directive is substituted with the definition of condition enum (requires -c --conditions option).

/*!stags:re2c ... */, /*!mtags:re2c ... */

These directives allow one to specify a template piece of code that is expanded for each s-tag/m-tag variable generated by re2c. This block has two optional configurations: format = "@@"; (specifies the template where @@ is substituted with the name of each tag variable), and separator = ""; (specifies the piece of code used to join the generated pieces for different tag variables).

/*!include:re2c FILE */

This directive allows one to include FILE (in the same sense as #include directive in C/C++).

/*!header:re2c:on*/

This directive marks the start of header file. Everything after it and up to the following /*!header:re2c:off*/ directive is processed by re2c and written to the header file specified with -t --type-header option.

/*!header:re2c:off*/

This directive marks the end of header file started with /*!header:re2c:on*/.

Configurations

re2c:flags:t, re2c:flags:type-header

Specify the name of the generated header file relative to the directory of the output file. (Same as -t, --type-header command-line option except that the filepath is relative.)

re2c:flags:input

Same as --input command-line option.

re2c:api:style

Allows one to specify the style of generic API. Possible values are functions and free-form. With functions style (the default for the C backend) API primitives behave like functions, and re2c generates parentheses with an argument list after the name of each primitive. With free-form style (the default for the Go backend) re2c treats API definitions as interpolated strings and substitutes argument placeholders with the actual argument values. This option can be overridden by options for individual API primitives, e.g. re2c:define:YYFILL:naked for YYFILL.

re2c:api:sigil

Allows one to specify the “sigil” symbol (or string) that is used to recognize argument placeholders in the definitions of generic API primitives. The default value is @@. Placeholders start with sigil, followed by the argument name in curly braces. For example, if sigil is set to $, then placeholders will have the form ${name}. Single-argument APIs may use shorthand notation without the name in braces. This option can be overridden by options for individual API primitives, e.g. re2c:define:YYFILL@len for YYFILL.

re2c:define:YYCTYPE

Defines YYCTYPE (see the user interface section).

re2c:define:YYCURSOR

Defines C API primitive YYCURSOR (see the user interface section).

re2c:define:YYLIMIT

Defines C API primitive YYLIMIT (see the user interface section).

re2c:define:YYMARKER

Defines C API primitive YYMARKER (see the user interface section).

re2c:define:YYCTXMARKER

Defines C API primitive YYCTXMARKER (see the user interface section).

re2c:define:YYFILL

Defines API primitive YYFILL (see the user interface section).

re2c:define:YYFILL@len

Specifies the sigil used for argument substitution in YYFILL definition. Defaults to @@. Overrides the more generic re2c:api:sigil configuration.

re2c:define:YYFILL:naked

Allows one to override re2c:api:style for YYFILL. Value 0 corresponds to free-form API style.

re2c:yyfill:enable

Defaults to 1 (YYFILL is enabled). Set this to zero to suppress the generation of YYFILL. Use warnings (-W option) and re2c:sentinel configuration to verify that the generated lexer cannot read past the end of input, as this might introduce severe security issues to your programs.

re2c:yyfill:parameter

Controls the argument in the parentheses that follow YYFILL. Defaults to 1, which means that the argument is generated. If zero, the argument is omitted. Can be overridden with re2c:define:YYFILL:naked or re2c:api:style.

re2c:eof

Specifies the sentinel symbol used with EOF rule $ to check for the end of input in the generated lexer. The default value is -1 (EOF rule is not used). Other possible values include all valid code units. Only decimal numbers are recognized.

re2c:sentinel

Specifies the sentinel symbol used with the sentinel method of checking for the end of input in the generated lexer (the case when bounds checking is disabled with re2c:yyfill:enable = 0; and EOF rule $ is not used). This configuration does not affect code generation. It is used by re2c to verify that the sentinel symbol is not allowed in the middle of the rule, and prevent possible reads past the end of buffer in the generated lexer. The default value is -1 (re2c assumes that the sentinel symbol is 0, which is the most common case). Other possible values include all valid code units. Only decimal numbers are recognized.

re2c:define:YYLESSTHAN

Defines generic API primitive YYLESSTHAN (see the user interface section).

re2c:yyfill:check

Setting this to zero allows to suppress the generation of YYFILL check (YYLESSTHAN in generic API of YYLIMIT-based comparison in default C API). This configuration is useful when the necessary input is always available. it defaults to 1 (the check is generated).

re2c:label:yyFillLabel

Allows one to change the prefix of YYFILL labels (used with EOF rule or with storable states).

re2c:define:YYPEEK

Defines generic API primitive YYPEEK (see the user interface section).

re2c:define:YYSKIP

Defines generic API primitive YYSKIP (see the user interface section).

re2c:define:YYBACKUP

Defines generic API primitive YYBACKUP (see the user interface section).

re2c:define:YYBACKUPCTX

Defines generic API primitive YYBACKUPCTX (see the user interface section).

re2c:define:YYRESTORE

Defines generic API primitive YYRESTORE (see the user interface section).

re2c:define:YYRESTORECTX

Defines generic API primitive YYRESTORECTX (see the user interface section).

re2c:define:YYRESTORETAG

Defines generic API primitive YYRESTORETAG (see the user interface section).

re2c:define:YYSHIFT

Defines generic API primitive YYSHIFT (see the user interface section).

re2c:define:YYSHIFTMTAG

Defines generic API primitive YYSHIFTMTAG (see the user interface section).

re2c:define:YYSHIFTSTAG

Defines generic API primitive YYSHIFTSTAG (see the user interface section).

re2c:define:YYSTAGN

Defines generic API primitive YYSTAGN (see the user interface section).

re2c:define:YYSTAGP

Defines generic API primitive YYSTAGP (see the user interface section).

re2c:define:YYMTAGN

Defines generic API primitive YYMTAGN (see the user interface section).

re2c:define:YYMTAGP

Defines generic API primitive YYMTAGP (see the user interface section).

re2c:flags:T, re2c:flags:tags

Same as -T --tags command-line option.

re2c:flags:P, re2c:flags:posix-captures

Same as -P --posix-captures command-line option.

re2c:tags:expression

Allows one to customize the way re2c addresses tag variables. By default re2c generates expressions of the form yyt<N>. This might be inconvenient, for example if tag variables are defined as fields in a struct. Re2c recognizes placeholder of the form @@{tag} or @@ and replaces it with the actual tag name. Sigil @@ can be redefined with re2c:api:sigil configuration. For example, setting re2c:tags:expression = "p->@@"; results in expressions of the form p->yyt<N> in the generated code.

re2c:tags:prefix

Allows one to override the prefix of tag variables (defaults to yyt).

re2c:flags:lookahead

Same as inverted --no-lookahead command-line option.

re2c:flags:optimize-tags

Same as inverted --no-optimize-tags command-line option.

re2c:define:YYCONDTYPE

Defines YYCONDTYPE (see the user interface section).

re2c:define:YYGETCONDITION

Defines API primitive YYGETCONDITION (see the user interface section).

re2c:define:YYGETCONDITION:naked

Allows one to override re2c:api:style for YYGETCONDITION. Value 0 corresponds to free-form API style.

re2c:define:YYSETCONDITION

Defines API primitive YYSETCONDITION (see the user interface section).

re2c:define:YYSETCONDITION@cond

Specifies the sigil used for argument substitution in YYSETCONDITION definition. The default value is @@. Overrides the more generic re2c:api:sigil configuration.

re2c:define:YYSETCONDITION:naked

Allows one to override re2c:api:style for YYSETCONDITION. Value 0 corresponds to free-form API style.

re2c:cond:goto

Allows one to customize the goto statements used with the shortcut :=> rules in conditions. The default value is goto @@;. Placeholders are substituted with condition name (see re2c:api;sigil and re2c:cond:goto@cond).

re2c:cond:goto@cond

Specifies the sigil used for argument substitution in re2c:cond:goto definition. The default value is @@. Overrides the more generic re2c:api:sigil configuration.

re2c:cond:divider

Defines the divider for condition blocks. The default value is /* *********************************** */. Placeholders are substituted with condition name (see re2c:api;sigil and re2c:cond:divider@cond).

re2c:cond:divider@cond

Specifies the sigil used for argument substitution in re2c:cond:divider definition. The default value is @@. Overrides the more generic re2c:api:sigil configuration.

re2c:condprefix

Specifies the prefix used for condition labels. The default value is yyc_.

re2c:condenumprefix

Specifies the prefix used for condition identifiers. The default value is yyc.

re2c:define:YYGETSTATE

Defines API primitive YYGETSTATE (see the user interface section).

re2c:define:YYGETSTATE:naked

Allows one to override re2c:api:style for YYGETSTATE. Value 0 corresponds to free-form API style.

re2c:define:YYSETSTATE

Defines API primitive YYSETSTATE (see the user interface section).

re2c:define:YYSETSTATE@state

Specifies the sigil used for argument substitution in YYSETSTATE definition. The default value is @@. Overrides the more generic re2c:api:sigil configuration.

re2c:define:YYSETSTATE:naked

Allows one to override re2c:api:style for YYSETSTATE. Value 0 corresponds to free-form API style.

re2c:state:abort

If set to a positive integer value, changes the form of the YYGETSTATE switch: instead of using default case to jump to the beginning of the lexer block, a -1 case is used, and the default case aborts the program.

re2c:state:nextlabel

With storable states, allows to control if the YYGETSTATE block is followed by a yyNext label (the default value is zero, which corresponds to no label). Instead of using yyNext it is possible to use re2c:startlabel to force the generation of a specific start label. Instead of using labels it is often more convenient to generate YYGETSTATE code using /*!getstate:re2c*/.

re2c:label:yyNext

Allows one to change the name of the yyNext label.

re2c:startlabel

Controls the generation of start label for the next lexer block. The default value is zero, which means that the start label is generated only if it is used. An integer value greater than zero forces the generation of start label even if it is unused by the lexer. A string value also forces start label generation and sets the label name to the specified string. This configuration applies only to the current block (it is reset to default for the next block).

re2c:flags:s, re2c:flags:nested-ifs

Same as -s --nested-ifs command-line option.

re2c:flags:b, re2c:flags:bit-vectors

Same as -b --bit-vectors command-line option.

re2c:variable:yybm

Overrides the name of the yybm variable.

re2c:yybm:hex

Defaults to zero (a decimal bitmap table is generated). If set to nonzero, a hexadecimal table is generated.

re2c:flags:g, re2c:flags:computed-gotos

Same as -g --computed-gotos command-line option.

re2c:cgoto:threshold

With -g --computed-gotos option this value specifies the complexity threshold that triggers the generation of jump tables instead of nested if statements and bitmaps. The default value is 9.

re2c:flags:case-ranges

Same as --case-ranges command-line option.

re2c:flags:e, re2c:flags:ecb

Same as -e --ecb command-line option.

re2c:flags:8, re2c:flags:utf-8

Same as -8 --utf-8 command-line option.

re2c:flags:w, re2c:flags:wide-chars

Same as -w --wide-chars command-line option.

re2c:flags:x, re2c:flags:utf-16

Same as -x --utf-16 command-line option.

re2c:flags:u, re2c:flags:unicode

Same as -u --unicode command-line option.

re2c:flags:encoding-policy

Same as --encoding-policy command-line option.

re2c:flags:empty-class

Same as --empty-class command-line option.

re2c:flags:case-insensitive

Same as --case-insensitive command-line option.

re2c:flags:case-inverted

Same as --case-inverted command-line option.

re2c:flags:i, re2c:flags:no-debug-info

Same as -i --no-debug-info command-line option.

re2c:indent:string

Specifies the string to use for indentation. The default value is "\t". Indent string should contain only whitespace characters. To disable indentation entirely, set this configuration to empty string "".

re2c:indent:top

Specifies the minimum amount of indentation to use. The default value is zero. The value should be a non-negative integer number.

re2c:labelprefix

Allows one to change the prefix of DFA state labels. The default value is yy.

re2c:yych:emit

Set this to zero to suppress the generation of yych definition. Defaults to 1 (the definition is generated).

re2c:variable:yych

Overrides the name of the yych variable.

re2c:yych:conversion

If set to nonzero, re2c automatically generates a cast to YYCTYPE every time yych is read. Defaults to zero (no cast).

re2c:variable:yyaccept

Overrides the name of the yyaccept variable.

re2c:variable:yytarget

Overrides the name of the yytarget variable.

re2c:variable:yystable

Deprecated.

re2c:variable:yyctable

When both -c --conditions and -g --computed-gotos are active, re2c will use this variable to generate a static jump table for YYGETCONDITION.

re2c:define:YYDEBUG

Defines YYDEBUG (see the user interface section).

re2c:flags:d, re2c:flags:debug-output

Same as -d --debug-output command-line option.

re2c:flags:dfa-minimization

Same as --dfa-minimization command-line option.

re2c:flags:eager-skip

Same as --eager-skip command-line option.

Regular expressions

re2c uses the following syntax for regular expressions:

  • "foo" case-sensitive string literal

  • 'foo' case-insensitive string literal

  • [a-xyz], [^a-xyz] character class (possibly negated)

  • . any character except newline

  • R \ S difference of character classes R and S

  • R* zero or more occurrences of R

  • R+ one or more occurrences of R

  • R? optional R

  • R{n} repetition of R exactly n times

  • R{n,} repetition of R at least n times

  • R{n,m} repetition of R from n to m times

  • (R) just R; parentheses are used to override precedence or for POSIX-style submatch

  • R S concatenation: R followed by S

  • R | S alternative: R or S

  • R / S lookahead: R followed by S, but S is not consumed

  • name the regular expression defined as name (or literal string "name" in Flex compatibility mode)

  • {name} the regular expression defined as name in Flex compatibility mode

  • @stag an s-tag: saves the last input position at which @stag matches in a variable named stag

  • #mtag an m-tag: saves all input positions at which #mtag matches in a variable named mtag

Character classes and string literals may contain the following escape sequences: \a, \b, \f, \n, \r, \t, \v, \\, octal escapes \ooo and hexadecimal escapes \xhh, \uhhhh and \Uhhhhhhhh.

EOF handling

Re2c provides a number of ways to handle end-of-input situation. Which way to use depends on the complexity of regular expressions, performance considerations, the need for input buffering and various other factors. EOF handling is probably the most complex part of re2c user interface — it definitely requires a bit of understanding of how the generated lexer works. But in return is allows the user to customize lexer for a particular environment and avoid the unnecessary overhead of generic methods when a simpler method is sufficient. Roughly speaking, there are four main methods:

  • using sentinel symbol (simple and efficient, but limited)

  • bounds checking with padding (generic, but complex)

  • EOF rule: a combination of sentinel symbol and bounds checking (generic and simple, can be more or less efficient than bounds checking with padding depending on the grammar)

  • using generic API (user-defined, so may be incorrect ;])

Using sentinel symbol

This is the simplest and the most efficient method. It is applicable in cases when the input is small enough to fit into a continuous memory buffer and there is a natural “sentinel” symbol — a code unit that is not allowed by any of the regular expressions in grammar (except possibly as a terminating character). Sentinel symbol never appears in well-formed input, therefore it can be appended at the end of input and used as a stop signal by the lexer. A good example of such input is a null-terminated C-string, provided that the grammar does not allow NULL in the middle of lexemes. Sentinel method is very efficient, because the lexer does not need to perform any additional checks for the end of input — it comes naturally as a part of processing the next character. It is very important that the sentinel symbol is not allowed in the middle of the rule — otherwise on some inputs the lexer may read past the end of buffer and crash or cause memory corruption. Re2c verifies this automatically. Use re2c:sentinel configuration to specify which sentinel symbol is used.

Below is an example of using sentinel method. Configuration re2c:yyfill:enable = 0; suppresses generation of end-of-input checks and YYFILL calls.

// re2c $INPUT -o $OUTPUT 
#include <assert.h>

// expect a null-terminated string
static int lex(const char *YYCURSOR)
{
    int count = 0;
loop:
    /*!re2c
    re2c:define:YYCTYPE = char;
    re2c:yyfill:enable = 0;

    *      { return -1; }
    [\x00] { return count; }
    [a-z]+ { ++count; goto loop; }
    [ ]+   { goto loop; }

    */
}

int main()
{
    assert(lex("") == 0);
    assert(lex("one two three") == 3);
    assert(lex("f0ur") == -1);
    return 0;
}

Bounds checking with padding

Bounds checking is a generic method: it can be used with any input grammar. The basic idea is simple: we need to check for the end of input before reading the next input character. However, if implemented in a straightforward way, this would be quite inefficient: checking on each input character would cause a major slowdown. Re2c avoids slowdown by generating checks only in certain key states of the lexer, and letting it run without checks in-between the key states. More precisely, re2c computes strongly connected components (SCCs) of the underlying DFA (which roughly correspond to loops), and generates only a few checks per each SCC (usually just one, but in general enough to make the SCC acyclic). The check is of the form (YYLIMIT - YYCURSOR) < n, where n is the maximal length of a simple path in the corresponding SCC. If this condiiton is true, the lexer calls YYFILL(n), which must either supply at least n input characters, or do not return. When the lexer continues after the check, it is certain that the next n characters can be read safely without checks.

This approach reduces the number of checks significantly (and makes the lexer much faster as a result), but it has a downside. Since the lexer checks for multiple characters at once, it may end up in a situation when there are a few remaining input characters (less than n) corresponding to a short path in the SCC, but the lexer cannot proceed because of the check, and YYFILL cannot supply more character because it is the end of input. To solve this problem, re2c requires that additional padding consisting of fake characters is appended at the end of input. The length of padding should be YYMAXFILL, which equals to the maximum n parameter to YYFILL and must be generated by re2c using /*!max:re2c*/ directive. The fake characters should not form a valid lexeme suffix, otherwise the lexer may be fooled into matching a fake lexeme. Usually it’s a good idea to use NULL characters for padding.

Below is an example of using bounds checking with padding. Note that the grammar rule for single-quoted strings allows arbitrary symbols in the middle of lexeme, so there is no natural sentinel in the grammar. Strings like "aha\0ha" are perfectly valid, but ill-formed strings like "aha\0 are also possible and shouldn’t crash the lexer. In this example we do not use buffer refilling, therefore YYFILL definition simply returns an error. Note that YYFILL will only be called after the lexer reaches padding, because only then will the check condition be satisfied.

// re2c $INPUT -o $OUTPUT 
#include <assert.h>
#include <stdlib.h>
#include <string.h>

/*!max:re2c*/

// expect YYMAXFILL-padded string
static int lex(const char *str, unsigned int len)
{
    const char *YYCURSOR = str, *YYLIMIT = str + len + YYMAXFILL;
    int count = 0;

loop:
    /*!re2c
    re2c:api:style = free-form;
    re2c:define:YYCTYPE = char;
    re2c:define:YYFILL = "return -1;";

    *                           { return -1; }
    [\x00]                      { return YYCURSOR == YYLIMIT ? count : -1; }
    ['] ([^'\\] | [\\][^])* ['] { ++count; goto loop; }
    [ ]+                        { goto loop; }

    */
}

// make a copy of the string with YYMAXFILL zeroes at the end
static void test(const char *str, unsigned int len, int res)
{
    char *s = (char*) malloc(len + YYMAXFILL);
    memcpy(s, str, len);
    memset(s + len, 0, YYMAXFILL);
    int r = lex(s, len);
    free(s);
    assert(r == res);
}

#define TEST(s, r) test(s, sizeof(s) - 1, r)
int main()
{
    TEST("", 0);
    TEST("'qu\0tes' 'are' 'fine: \\'' ", 3);
    TEST("'unterminated\\'", -1);
    return 0;
}

EOF rule

EOF rule $ was introduced in version 1.2. It is a hybrid approach that tries to take the best of both worlds: simplicity and efficiency of the sentinel method combined with the generality of bounds-checking method. The idea is to appoint an arbitrary symbol to be the sentinel, and only perform further bounds checking if the sentinel symbol matches (more precisely, if the symbol class that contains it matches). The check is of the form YYLIMIT <= YYCURSOR. If this condition is not satisfied, then the sentinel is just an ordinary input character and the lexer continues. Otherwise this is a real sentinel, and the lexer calls YYFILL(). If YYFILL returns zero, the lexer assumes that it has more input and tries to re-match. Otherwise YYFILL returns non-zero and the lexer knows that it has reached the end of input. At this point there are three possibilities. First, it might have already matched a shorter lexeme — in this case it just rolls back to the last accepting state. Second, it might have consumed some characters, but failed to match — in this case it falls back to default rule *. Finally, it might be in the initial state — in this (and only this!) case it matches EOF rule $.

Below is an example of using EOF rule. Configuration re2c:yyfill:enable = 0; suppresses generation of YYFILL calls (but not the bounds checks).

// re2c $INPUT -o $OUTPUT 
#include <assert.h>

// expect a null-terminated string
static int lex(const char *str, unsigned int len)
{
    const char *YYCURSOR = str, *YYLIMIT = str + len, *YYMARKER;
    int count = 0;

loop:
    /*!re2c
    re2c:define:YYCTYPE = char;
    re2c:yyfill:enable = 0;
    re2c:eof = 0;

    *                           { return -1; }
    $                           { return count; }
    ['] ([^'\\] | [\\][^])* ['] { ++count; goto loop; }
    [ ]+                        { goto loop; }

    */
}

#define TEST(s, r) assert(lex(s, sizeof(s) - 1) == r)
int main()
{
    TEST("", 0);
    TEST("'qu\0tes' 'are' 'fine: \\'' ", 3);
    TEST("'unterminated\\'", -1);
    return 0;
}

Using generic API

Generic API can be used with any of the above methods. It also allows one to use a user-defined method by placing EOF checks in one of the basic primitives. Usually this is either YYSKIP (the check is performed when advancing to the next input character), or YYPEEK (the check is performed when reading the next input character). The resulting methods are inefficient, as they check on each input character. However, they can be useful in cases when the input cannot be buffered or padded and does not contain a sentinel character at the end. One should be cautious when using such ad-hoc methods, as it is easy to overlook some corner cases and come up with a method that only partially works. Also it should be noted that not everything can be expressed via generic API: for example, it is impossible to reimplement the way EOF rule works (in particular, it is impossible to re-match the character after successful YYFILL).

Below is an example of using YYSKIP to perform bounds checking without padding. YYFILL generation is suppressed using re2c:yyfill:enable = 0; configuration. Note that if the grammar was more complex, this method might not work in case when two rules overlap and EOF check fails after a shorter lexeme has already been matched (as it happens in our example, there are no overlapping rules).

// re2c $INPUT -o $OUTPUT 
#include <assert.h>
#include <stdlib.h>
#include <string.h>

// expect a string without terminating null
static int lex(const char *str, unsigned int len)
{
    const char *cur = str, *lim = str + len, *mar;
    int count = 0;

loop:
    /*!re2c
    re2c:yyfill:enable = 0;
    re2c:eof = 0;
    re2c:flags:input = custom;
    re2c:api:style = free-form;
    re2c:define:YYCTYPE    = char;
    re2c:define:YYLESSTHAN = "cur >= lim";
    re2c:define:YYPEEK     = "cur < lim ? *cur : 0";  // fake null
    re2c:define:YYSKIP     = "++cur;";
    re2c:define:YYBACKUP   = "mar = cur;";
    re2c:define:YYRESTORE  = "cur = mar;";

    *                           { return -1; }
    $                           { return count; }
    ['] ([^'\\] | [\\][^])* ['] { ++count; goto loop; }
    [ ]+                        { goto loop; }

    */
}

// make a copy of the string without terminating null
static void test(const char *str, unsigned int len, int res)
{
    char *s = (char*) malloc(len);
    memcpy(s, str, len);
    int r = lex(s, len);
    free(s);
    assert(r == res);
}

#define TEST(s, r) test(s, sizeof(s) - 1, r)
int main()
{
    TEST("", 0);
    TEST("'qu\0tes' 'are' 'fine: \\'' ", 3);
    TEST("'unterminated\\'", -1);
    return 0;
}

Buffer refilling

The need for buffering arises when the input cannot be mapped in memory all at once: either it is too large, or it comes in a streaming fashion (like reading from a socket). The usual technique in such cases is to allocate a fixed-sized memory buffer and process input in chunks that fit into the buffer. When the current chunk is processed, it is moved out and new data is moved in. In practice it is somewhat more complex, because lexer state consists not of a single input position, but a set of interrelated posiitons:

  • cursor: the next input character to be read (YYCURSOR in default API or YYSKIP/YYPEEK in generic API)

  • limit: the position after the last available input character (YYLIMIT in default API, implicitly handled by YYLESSTHAN in generic API)

  • marker: the position of the most recent match, if any (YYMARKER in default API or YYBACKUP/YYRESTORE in generic API)

  • token: the start of the current lexeme (implicit in re2c API, as it is not needed for the normal lexer operation and can be defined and updated by the user)

  • context marker: the position of the trailing context (YYCTXMARKER in default API or YYBACKUPCTX/YYRESTORECTX in generic API)

  • tag variables: submatch positions (defined with /*!stags:re2c*/ and /*!mtags:re2c*/ directives and YYSTAGP/YYSTAGN/YYMTAGP/YYMTAGN in generic API)

Not all these are used in every case, but if used, they must be updated by YYFILL. All active positions are contained in the segment between token and cursor, therefore everything between buffer start and token can be discarded, the segment from token and up to limit should be moved to the beginning of buffer, and the free space at the end of buffer should be filled with new data. In order to avoid frequent YYFILL calls it is best to fill in as many input characters as possible (even though fewer characters might suffice to resume the lexer). The details of YYFILL implementation are slightly different depending on which EOF handling method is used: the case of EOF rule is somewhat simpler than the case of bounds-checking with padding. Also note that if -f --storable-state option is used, YYFILL has slightly different semantics (desrbed in the section about storable state).

YYFILL with EOF rule

If EOF rule is used, YYFILL is a function-like primitive that accepts no arguments and returns a value which is checked against zero. YYFILL invocation is triggered by condition YYLIMIT <= YYCURSOR in default API and YYLESSTHAN() in generic API. A non-zero return value means that YYFILL has failed. A successful YYFILL call must supply at least one character and adjust input positions accordingly. Limit must always be set to one after the last input position in buffer, and the character at the limit position must be the sentinel symbol specified by re2c:eof configuration. The pictures below show the relative locations of input positions in buffer before and after YYFILL call (sentinel symbol is marked with #, and the second picture shows the case when there is not enough input to fill the whole buffer).

               <-- shift -->
             >-A------------B---------C-------------D#-----------E->
             buffer       token    marker         limit,
                                                  cursor
>-A------------B---------C-------------D------------E#->
             buffer,  marker        cursor        limit
             token

               <-- shift -->
             >-A------------B---------C-------------D#--E (EOF)
             buffer       token    marker         limit,
                                                  cursor
>-A------------B---------C-------------D---E#........
             buffer,  marker       cursor limit
             token

Here is an example of a program that reads input file input.txt in chunks of 4096 bytes and uses EOF rule.

// re2c $INPUT -o $OUTPUT 
#include <assert.h>
#include <stdio.h>
#include <string.h>

#define SIZE 4096

typedef struct {
    FILE *file;
    char buf[SIZE + 1], *lim, *cur, *mar, *tok;
    int eof;
} Input;

static int fill(Input *in)
{
    if (in->eof) {
        return 1;
    }
    const size_t free = in->tok - in->buf;
    if (free < 1) {
        return 2;
    }
    memmove(in->buf, in->tok, in->lim - in->tok);
    in->lim -= free;
    in->cur -= free;
    in->mar -= free;
    in->tok -= free;
    in->lim += fread(in->lim, 1, free, in->file);
    in->lim[0] = 0;
    in->eof |= in->lim < in->buf + SIZE;
    return 0;
}

static void init(Input *in, FILE *file)
{
    in->file = file;
    in->cur = in->mar = in->tok = in->lim = in->buf + SIZE;
    in->eof = 0;
    fill(in);
}

static int lex(Input *in)
{
    int count = 0;
loop:
    in->tok = in->cur;
    /*!re2c
    re2c:eof = 0;
    re2c:api:style = free-form;
    re2c:define:YYCTYPE  = char;
    re2c:define:YYCURSOR = in->cur;
    re2c:define:YYMARKER = in->mar;
    re2c:define:YYLIMIT  = in->lim;
    re2c:define:YYFILL   = "fill(in) == 0";

    *                           { return -1; }
    $                           { return count; }
    ['] ([^'\\] | [\\][^])* ['] { ++count; goto loop; }
    [ ]+                        { goto loop; }

    */
}

int main()
{
    const char *fname = "input";
    const char str[] = "'qu\0tes' 'are' 'fine: \\'' ";
    FILE *f;
    Input in;

    // prepare input file: a few times the size of the buffer,
    // containing strings with zeroes and escaped quotes
    f = fopen(fname, "w");
    for (int i = 0; i < SIZE; ++i) {
        fwrite(str, 1, sizeof(str) - 1, f);
    }
    fclose(f);

    f = fopen(fname, "r");
    init(&in, f);
    assert(lex(&in) == SIZE * 3);
    fclose(f);

    remove(fname);
    return 0;
}

YYFILL with padding

In the default case (when EOF rule is not used) YYFILL is a function-like primitive that accepts a single argument and does not return any value. YYFILL invocation is triggered by condition (YYLIMIT - YYCURSOR) < n in default API and YYLESSTHAN(n) in generic API. The argument passed to YYFILL is the minimal number of characters that must be supplied. If it fails to do so, YYFILL must not return to the lexer (for that reason it is best implemented as a macro that returns from the calling function on failure). In case of a successful YYFILL invocation the limit position must be set either to one after the last input position in buffer, or to the end of YYMAXFILL padding (in case YYFILL has successfully read at least n characters, but not enough to fill the entire buffer). The pictures below show the relative locations of input positions in buffer before and after YYFILL invocation (YYMAXFILL padding on the second picture is marked with # symbols).

               <-- shift -->                 <-- need -->
             >-A------------B---------C-----D-------E---F--------G->
             buffer       token    marker cursor  limit

>-A------------B---------C-----D-------E---F--------G->
             buffer,  marker cursor               limit
             token

               <-- shift -->                 <-- need -->
             >-A------------B---------C-----D-------E-F        (EOF)
             buffer       token    marker cursor  limit

>-A------------B---------C-----D-------E-F###############
             buffer,  marker cursor                   limit
             token                        <- YYMAXFILL ->

Here is an example of a program that reads input file input.txt in chunks of 4096 bytes and uses bounds-checking with padding.

// re2c $INPUT -o $OUTPUT 
#include <assert.h>
#include <stdio.h>
#include <string.h>

/*!max:re2c*/
#define SIZE 4096

typedef struct {
    FILE *file;
    char buf[SIZE + YYMAXFILL], *lim, *cur, *mar, *tok;
    int eof;
} Input;

static int fill(Input *in, size_t need)
{
    if (in->eof) {
        return 1;
    }
    const size_t free = in->tok - in->buf;
    if (free < need) {
        return 2;
    }
    memmove(in->buf, in->tok, in->lim - in->tok);
    in->lim -= free;
    in->cur -= free;
    in->mar -= free;
    in->tok -= free;
    in->lim += fread(in->lim, 1, free, in->file);
    if (in->lim < in->buf + SIZE) {
        in->eof = 1;
        memset(in->lim, 0, YYMAXFILL);
        in->lim += YYMAXFILL;
    }
    return 0;
}

static void init(Input *in, FILE *file)
{
    in->file = file;
    in->cur = in->mar = in->tok = in->lim = in->buf + SIZE;
    in->eof = 0;
    fill(in, 1);
}

static int lex(Input *in)
{
    int count = 0;
loop:
    in->tok = in->cur;
    /*!re2c
    re2c:api:style = free-form;
    re2c:define:YYCTYPE  = char;
    re2c:define:YYCURSOR = in->cur;
    re2c:define:YYMARKER = in->mar;
    re2c:define:YYLIMIT  = in->lim;
    re2c:define:YYFILL   = "if (fill(in, @@) != 0) return -1;";

    *                           { return -1; }
    [\x00]                      { return (YYMAXFILL == in->lim - in->tok) ? count : -1; }
    ['] ([^'\\] | [\\][^])* ['] { ++count; goto loop; }
    [ ]+                        { goto loop; }

    */
}

int main()
{
    const char *fname = "input";
    const char str[] = "'qu\0tes' 'are' 'fine: \\'' ";
    FILE *f;
    Input in;

    // prepare input file: a few times the size of the buffer,
    // containing strings with zeroes and escaped quotes
    f = fopen(fname, "w");
    for (int i = 0; i < SIZE; ++i) {
        fwrite(str, 1, sizeof(str) - 1, f);
    }
    fclose(f);

    f = fopen(fname, "r");
    init(&in, f);
    assert(lex(&in) == SIZE * 3);
    fclose(f);

    remove(fname);
    return 0;
}

Include files

Re2c allows one to include other files using directive /*!include:re2c FILE */, where FILE is the name of file to be included. Re2c looks for included files in the directory of the including file and in include locations, which can be specified with -I option. Re2c include directive works in the same way as C/C++ #include: the contents of FILE are copy-pasted verbatim in place of the directive. Include files may have further includes of their own. Re2c provides some predefined include files that can be found in the include/ subdirectory of the project. These files contain definitions that can be useful to other projects (such as Unicode categories) and form something like a standard library for re2c. Here is an example:

Include file (definitions.h):

typedef enum { OK, FAIL } Result;

/*!re2c
    number = [1-9][0-9]*;
*/

Input file:

// re2c $INPUT -o $OUTPUT -i
#include <assert.h>
/*!include:re2c "definitions.h" */

Result lex(const char *YYCURSOR)
{
    /*!re2c
    re2c:define:YYCTYPE = char;
    re2c:yyfill:enable = 0;

    number { return OK; }
    *      { return FAIL; }
    */
}

int main()
{
    assert(lex("123") == OK);
    return 0;
}

Header files

Re2c allows one to generate header file from the input .re file using option -t, --type-header or configuration re2c:flags:type-header and directives /*!header:re2c:on*/ and /*!header:re2c:off*/. The first directive marks the beginning of header file, and the second directive marks the end of it. Everything between these directives is processed by re2c, and the generated code is written to the file specified by the -t --type-header option (or stdout if this option was not used). Autogenerated header file may be needed in cases when re2c is used to generate definitions of constants, variables and structs that must be visible from other translation units.

Here is an example of generating a header file that contains definition of the lexer state with tag variables (the number variables depends on the regular grammar and is unknown to the programmer).

// re2c $INPUT -o $OUTPUT -i --type-header src/lexer/lexer.h
#include <assert.h>
#include "src/lexer/lexer.h" // generated by re2c

/*!header:re2c:on*/

typedef struct {
    const char *str, *cur, *mar;
    /*!stags:re2c format = "const char *@@{tag}; "; */
} LexerState;

/*!header:re2c:off*/

int lex(LexerState *st)
{
    /*!re2c
    re2c:flags:type-header = "src/lexer/lexer.h";
    re2c:yyfill:enable = 0;
    re2c:flags:tags = 1;
    re2c:define:YYCTYPE  = char;
    re2c:define:YYCURSOR = "st->cur";
    re2c:define:YYMARKER = "st->mar";
    re2c:tags:expression = "st->@@{tag}";

    [x]{1,4} / [x]{3,5} { return 0; } // ambiguous trailing context
    *                   { return 1; }
    */
}

int main()
{
    LexerState st;
    st.str = st.cur = "xxxxxxxx";
    assert(lex(&st) == 0 && st.cur - st.str == 4);
    return 0;
}

The generated header file:

/* Generated by re2c */


typedef struct {
    const char *str, *cur, *mar;
    const char *yyt1; const char *yyt2; const char *yyt3; 
} LexerState;

Submatch extraction

Re2c has two options for submatch extraction.

The first option is -T --tags. With this option one can use standalone tags of the form @stag and #mtag, where stag and mtag are arbitrary used-defined names. Tags can be used anywhere inside of a regular expression; semantically they are just position markers. Tags of the form @stag are called s-tags: they denote a single submatch value (the last input position where this tag matched). Tags of the form #mtag are called m-tags: they denote multiple submatch values (the whole history of repetitions of this tag). All tags should be defined by the user as variables with the corresponding names. With standalone tags re2c uses leftmost greedy disambiguation: submatch positions correspond to the leftmost matching path through the regular expression.

The second option is -P --posix-captures: it enables POSIX-compliant capturing groups. In this mode parentheses in regular expressions denote the beginning and the end of capturing groups; the whole regular expression is group number zero. The number of groups for the matching rule is stored in a variable yynmatch, and submatch results are stored in yypmatch array. Both yynmatch and yypmatch should be defined by the user, and yypmatch size must be at least [yynmatch * 2]. Re2c provides a directive /*!maxnmatch:re2c*/ that defines YYMAXNMATCH: a constant equal to the maximal value of yynmatch among all rules. Note that re2c implements POSIX-compliant disambiguation: each subexpression matches as long as possible, and subexpressions that start earlier in regular expression have priority over those starting later. Capturing groups are translated into s-tags under the hood, therefore we use the word “tag” to describe them as well.

With both -P --posix-captures and T --tags options re2c uses efficient submatch extraction algorithm described in the Tagged Deterministic Finite Automata with Lookahead paper. The overhead on submatch extraction in the generated lexer grows with the number of tags — if this number is moderate, the overhead is barely noticeable. In the lexer tags are implemented using a number of tag variables generated by re2c. There is no one-to-one correspondence between tag variables and tags: a single variable may be reused for different tags, and one tag may require multiple variables to hold all its ambiguous values. Eventually ambiguity is resolved, and only one final variable per tag survives. When a rule matches, all its tags are set to the values of the corresponding tag variables. The exact number of tag variables is unknown to the user; this number is determined by re2c. However, tag variables should be defined by the user as a part of the lexer state and updated by YYFILL, therefore re2c provides directives /*!stags:re2c*/ and /*!mtags:re2c*/ that can be used to declare, initialize and manipulate tag variables. These directives have two optional configurations: format = "@@"; (specifies the template where @@ is substituted with the name of each tag variable), and separator = ""; (specifies the piece of code used to join the generated pieces for different tag variables).

S-tags support the following operations:

  • save input position to an s-tag: t = YYCURSOR with default API or a user-defined operation YYSTAGP(t) with generic API

  • save default value to an s-tag: t = NULL with default API or a user-defined operation YYSTAGN(t) with generic API

  • copy one s-tag to another: t1 = t2

M-tags support the following operations:

  • append input position to an m-tag: a user-defined operation YYMTAGP(t) with both default and generic API

  • append default value to an m-tag: a user-defined operation YYMTAGN(t) with both default and generic API

  • copy one m-tag to another: t1 = t2

S-tags can be implemented as scalar values (pointers or offsets). M-tags need a more complex representation, as they need to store a sequence of tag values. The most naive and inefficient representation of an m-tag is a list (array, vector) of tag values; a more efficient representation is to store all m-tags in a prefix-tree represented as array of nodes (v, p), where v is tag value and p is a pointer to parent node.

Here is an example of using s-tags to parse an IPv4 address.

// re2c $INPUT -o $OUTPUT 
#include <assert.h>
#include <stdint.h>

static uint32_t num(const char *s, const char *e)
{
    uint32_t n = 0;
    for (; s < e; ++s) n = n * 10 + (*s - '0');
    return n;
}

static const uint64_t ERROR = ~0lu;

static uint64_t lex(const char *YYCURSOR)
{
    const char *YYMARKER, *o1, *o2, *o3, *o4;
    /*!stags:re2c format = 'const char *@@;'; */

    /*!re2c
    re2c:yyfill:enable = 0;
    re2c:flags:tags = 1;
    re2c:define:YYCTYPE = char;

    octet = [0-9] | [1-9][0-9] | [1][0-9][0-9] | [2][0-4][0-9] | [2][5][0-5];
    dot = [.];
    end = [\x00];

    @o1 octet dot @o2 octet dot @o3 octet dot @o4 octet end {
        return num(o4, YYCURSOR - 1)
            + (num(o3, o4 - 1) << 8)
            + (num(o2, o3 - 1) << 16)
            + (num(o1, o2 - 1) << 24);
    }
    * { return ERROR; }
    */
}

int main()
{
    assert(lex("1.2.3.4") == 0x01020304);
    assert(lex("127.0.0.1") == 0x7f000001);
    assert(lex("255.255.255.255") == 0xffffffff);
    assert(lex("1.2.3.") == ERROR);
    assert(lex("1.2.3.256") == ERROR);
    return 0;
}

Here is an example of using POSIX capturing groups to parse an IPv4 address.

// re2c $INPUT -o $OUTPUT 
#include <assert.h>
#include <stdint.h>

static uint32_t num(const char *s, const char *e)
{
    uint32_t n = 0;
    for (; s < e; ++s) n = n * 10 + (*s - '0');
    return n;
}

/*!maxnmatch:re2c*/
static const uint64_t ERROR = ~0lu;

static uint64_t lex(const char *YYCURSOR)
{
    const char *YYMARKER;
    const char *yypmatch[YYMAXNMATCH * 2];
    uint32_t yynmatch;
    /*!stags:re2c format = 'const char *@@;'; */

    /*!re2c
    re2c:yyfill:enable = 0;
    re2c:flags:posix-captures = 1;
    re2c:define:YYCTYPE = char;

    octet = [0-9] | [1-9][0-9] | [1][0-9][0-9] | [2][0-4][0-9] | [2][5][0-5];
    dot = [.];
    end = [\x00];

    (octet) dot (octet) dot (octet) dot (octet) end {
        assert(yynmatch == 5);
        return num(yypmatch[8], yypmatch[9])
            + (num(yypmatch[6], yypmatch[7]) << 8)
            + (num(yypmatch[4], yypmatch[5]) << 16)
            + (num(yypmatch[2], yypmatch[3]) << 24);
    }
    * { return ERROR; }
    */
}

int main()
{
    assert(lex("1.2.3.4") == 0x01020304);
    assert(lex("127.0.0.1") == 0x7f000001);
    assert(lex("255.255.255.255") == 0xffffffff);
    assert(lex("1.2.3.") == ERROR);
    assert(lex("1.2.3.256") == ERROR);
    return 0;
}

Here is an example of using m-tags to parse a semicolon-separated sequence of words (C++). Tag variables are stored in a tree that is packed in a vector.

// re2c $INPUT -o $OUTPUT 
#include <assert.h>
#include <vector>
#include <string>

static const int ROOT = -1;

struct Mtag {
    int pred;
    const char *tag;
};

typedef std::vector<Mtag> MtagTree;
typedef std::vector<std::string> Words;

static void mtag(int *pt, const char *t, MtagTree *tree)
{
    Mtag m = {*pt, t};
    *pt = (int)tree->size();
    tree->push_back(m);
}

static void unfold(const MtagTree &tree, int x, int y, Words &words)
{
    if (x == ROOT) return;
    unfold(tree, tree[x].pred, tree[y].pred, words);
    const char *px = tree[x].tag, *py = tree[y].tag;
    words.push_back(std::string(px, py - px));
}

#define YYMTAGP(t) mtag(&t, YYCURSOR, &tree)
#define YYMTAGN(t) mtag(&t, NULL,     &tree)
static bool lex(const char *YYCURSOR, Words &words)
{
    const char *YYMARKER;
    /*!mtags:re2c format = "int @@ = ROOT;"; */
    MtagTree tree;
    int x, y;

    /*!re2c
    re2c:yyfill:enable = 0;
    re2c:flags:tags = 1;
    re2c:define:YYCTYPE = char;

    (#x [a-z]+ #y [;])+ {
        words.clear();
        unfold(tree, x, y, words);
        return true;
    }
    * { return false; }
    */
}

int main()
{
    Words w;
    assert(lex("one;two;three;", w) && w == Words({"one", "two", "three"}));
    return 0;
}

Storable state

With -f --storable-state option re2c generates a lexer that can store its current state, return to the caller, and later resume operations exactly where it left off. The default mode of operation in re2c is a “pull” model, in which the lexer “pulls” more input whenever it needs it. This may be unacceptable in cases when the input becomes available piece by piece (for example, if the lexer is invoked by the parser, or if the lexer program communicates via a socket protocol with some other program that must wait for a reply from the lexer before it transmits the next message). Storable state feature is intended exactly for such cases: it allows one to generate lexers that work in a “push” model. When the lexer needs more input, it stores its state and returns to the caller. Later, when more input becomes available, the caller resumes the lexer exactly where it stopped. There are a few changes necessary compared to the “pull” model:

  • Define YYSETSTATE() and YYGETSTATE(state) promitives.

  • Define yych, yyaccept and state variables as a part of persistent lexer state. The state variable should be initialized to -1.

  • YYFILL should return to the outer program instead of trying to supply more input. Return code should indicate that lexer needs more input.

  • The outer program should recognize situations when lexer needs more input and respond appropriately.

  • Use /*!getstate:re2c*/ directive if it is necessary to execute any code before entering the lexer.

  • Use configurations state:abort and state:nextlabel to further tweak the generated code.

Here is an example of a “push”-model lexer that reads input from stdin and expects a sequence of words separated by spaces and newlines. The lexer loops forever, waiting for more input. It can be terminated by sending a special EOF token — a word “stop”, in which case the lexer terminates successfully and prints the number of words it has seen. Abnormal termination happens in case of a syntax error, premature end of input (without the “stop” word) or in case the buffer is too small to hold a lexeme (for example, if one of the words exceeds buffer size). Premature end of input happens in case the lexer fails to read any input while being in the initial state — this is the only case when EOF rule matches. Note that the lexer may call YYFILL twice before terminating (and thus require hitting Ctrl+D a few times). First time YYFILL is called when the lexer expects continuation of the current greedy lexeme (either a word or a whitespace sequence). If YYFILL fails, the lexer knows that it has reached the end of the current lexeme and executes the corresponding semantic action. The action jumps to the beginning of the loop, the lexer enters the initial state and calls YYFILL once more. If it fails, the lexer matches EOF rule. (Alternatively EOF rule can be used for termination instead of a special EOF lexeme.)

// re2c $INPUT -o $OUTPUT -f
#include <assert.h>
#include <stdio.h>
#include <string.h>

#define DEBUG    0
#define LOG(...) if (DEBUG) fprintf(stderr, __VA_ARGS__);
#define BUFSIZE  10

typedef struct {
    FILE *file;
    char buf[BUFSIZE + 1], *lim, *cur, *mar, *tok;
    unsigned yyaccept;
    int state;
} Input;

static void init(Input *in, FILE *f)
{
    in->file = f;
    in->cur = in->mar = in->tok = in->lim = in->buf + BUFSIZE;
    in->lim[0] = 0; // append sentinel symbol
    in->yyaccept = 0;
    in->state = -1;
}

typedef enum {END, READY, WAITING, BAD_PACKET, BIG_PACKET} Status;

static Status fill(Input *in)
{
    const size_t shift = in->tok - in->buf;
    const size_t free = BUFSIZE - (in->lim - in->tok);

    if (free < 1) return BIG_PACKET;

    memmove(in->buf, in->tok, BUFSIZE - shift);
    in->lim -= shift;
    in->cur -= shift;
    in->mar -= shift;
    in->tok -= shift;

    const size_t read = fread(in->lim, 1, free, in->file);
    in->lim += read;
    in->lim[0] = 0; // append sentinel symbol

    return READY;
}

static Status lex(Input *in, unsigned int *recv)
{
    char yych;
    /*!getstate:re2c*/
loop:
    in->tok = in->cur;
    /*!re2c
        re2c:eof = 0;
        re2c:api:style = free-form;
        re2c:define:YYCTYPE    = "char";
        re2c:define:YYCURSOR   = "in->cur";
        re2c:define:YYMARKER   = "in->mar";
        re2c:define:YYLIMIT    = "in->lim";
        re2c:define:YYGETSTATE = "in->state";
        re2c:define:YYSETSTATE = "in->state = @@;";
        re2c:define:YYFILL     = "return WAITING;";

        packet = [a-z]+[;];

        *      { return BAD_PACKET; }
        $      { return END; }
        packet { *recv = *recv + 1; goto loop; }
    */
}

void test(const char **packets, Status status)
{
    const char *fname = "pipe";
    FILE *fw = fopen(fname, "w");
    FILE *fr = fopen(fname, "r");
    setvbuf(fw, NULL, _IONBF, 0);
    setvbuf(fr, NULL, _IONBF, 0);

    Input in;
    init(&in, fr);
    Status st;
    unsigned int send = 0, recv = 0;

    for (;;) {
        st = lex(&in, &recv);
        if (st == END) {
            LOG("done: got %u packets\n", recv);
            break;
        } else if (st == WAITING) {
            LOG("waiting...\n");
            if (*packets) {
                LOG("sent packet %u\n", send);
                fprintf(fw, "%s", *packets++);
                ++send;
            }
            st = fill(&in);
            LOG("queue: '%s'\n", in.buf);
            if (st == BIG_PACKET) {
                LOG("error: packet too big\n");
                break;
            }
            assert(st == READY);
        } else {
            assert(st == BAD_PACKET);
            LOG("error: ill-formed packet\n");
            break;
        }
    }

    LOG("\n");
    assert(st == status);
    if (st == END) assert(recv == send);

    fclose(fw);
    fclose(fr);
    remove(fname);
}

int main()
{
    const char *packets1[] = {0};
    const char *packets2[] = {"zero;", "one;", "two;", "three;", "four;", 0};
    const char *packets3[] = {"zer0;", 0};
    const char *packets4[] = {"goooooooooogle;", 0};

    test(packets1, END);
    test(packets2, END);
    test(packets3, BAD_PACKET);
    test(packets4, BIG_PACKET);

    return 0;
}

Reusable blocks

Reuse mode is enabled with the -r --reusable option. In this mode re2c allows one to reuse definitions, configurations and rules specified by a /*!rules:re2c*/ block in subsequent /*!use:re2c*/ blocks. As of re2c-1.2 it is possible to mix such blocks with normal /*!re2c*/ blocks; prior to that re2c expects a single rules-block followed by use-blocks (normal blocks are disallowed). Use-blocks can have additional definitions, configurations and rules: they are merged to those specified by the rules-block. A very common use case for -r --reusable option is a lexer that supports multiple input encodings: lexer rules are defined once and reused multiple times with encoding-specific configurations, such as re2c:flags:utf-8.

Below is an example of a multi-encoding lexer: it reads a phrase with Unicode math symbols and accepts input either in UTF8 or in UT32. Note that the --input-encoding utf8 option allows us to write UTF8-encoded symbols in the regular expressions; without this option re2c would parse them as a plain ASCII byte sequnce (and we would have to use hexadecimal escape sequences).

// re2c $INPUT -o $OUTPUT -r --input-encoding utf8
#include <assert.h>
#include <stdint.h>

/*!rules:re2c
    re2c:yyfill:enable = 0;

    "∀x ∃y: p(x, y)" { return 0; }
    *                { return 1; }
*/

static int lex_utf8(const uint8_t *YYCURSOR)
{
    const uint8_t *YYMARKER;
    /*!use:re2c
    re2c:define:YYCTYPE = uint8_t;
    re2c:flags:8 = 1;
    */
}

static int lex_utf32(const uint32_t *YYCURSOR)
{
    const uint32_t *YYMARKER;
    /*!use:re2c
    re2c:define:YYCTYPE = uint32_t;
    re2c:flags:8 = 0;
    re2c:flags:u = 1;
    */
}

int main()
{
    static const uint8_t s8[] = // UTF-8
        { 0xe2, 0x88, 0x80, 0x78, 0x20, 0xe2, 0x88, 0x83, 0x79
        , 0x3a, 0x20, 0x70, 0x28, 0x78, 0x2c, 0x20, 0x79, 0x29 };

    static const uint32_t s32[] = // UTF32
        { 0x00002200, 0x00000078, 0x00000020, 0x00002203
        , 0x00000079, 0x0000003a, 0x00000020, 0x00000070
        , 0x00000028, 0x00000078, 0x0000002c, 0x00000020
        , 0x00000079, 0x00000029 };

    assert(lex_utf8(s8) == 0);
    assert(lex_utf32(s32) == 0);
    return 0;
}

Encoding support

re2c supports the following encodings: ASCII (default), EBCDIC (-e), UCS-2 (-w), UTF-16 (-x), UTF-32 (-u) and UTF-8 (-8). See also inplace configuration re2c:flags.

The following concepts should be clarified when talking about encodings. A code point is an abstract number that represents a single symbol. A code unit is the smallest unit of memory, which is used in the encoded text (it corresponds to one character in the input stream). One or more code units may be needed to represent a single code point, depending on the encoding. In a fixed-length encoding, each code point is represented with an equal number of code units. In variable-length encodings, different code points can be represented with different number of code units.

  • ASCII is a fixed-length encoding. Its code space includes 0x100 code points, from 0 to 0xFF. A code point is represented with exactly one 1-byte code unit, which has the same value as the code point. The size of YYCTYPE must be 1 byte.

  • EBCDIC is a fixed-length encoding. Its code space includes 0x100 code points, from 0 to 0xFF. A code point is represented with exactly one 1-byte code unit, which has the same value as the code point. The size of YYCTYPE must be 1 byte.

  • UCS-2 is a fixed-length encoding. Its code space includes 0x10000 code points, from 0 to 0xFFFF. One code point is represented with exactly one 2-byte code unit, which has the same value as the code point. The size of YYCTYPE must be 2 bytes.

  • UTF-16 is a variable-length encoding. Its code space includes all Unicode code points, from 0 to 0xD7FF and from 0xE000 to 0x10FFFF. One code point is represented with one or two 2-byte code units. The size of YYCTYPE must be 2 bytes.

  • UTF-32 is a fixed-length encoding. Its code space includes all Unicode code points, from 0 to 0xD7FF and from 0xE000 to 0x10FFFF. One code point is represented with exactly one 4-byte code unit. The size of YYCTYPE must be 4 bytes.

  • UTF-8 is a variable-length encoding. Its code space includes all Unicode code points, from 0 to 0xD7FF and from 0xE000 to 0x10FFFF. One code point is represented with a sequence of one, two, three, or four 1-byte code units. The size of YYCTYPE must be 1 byte.

In Unicode, values from range 0xD800 to 0xDFFF (surrogates) are not valid Unicode code points. Any encoded sequence of code units that would map to Unicode code points in the range 0xD800-0xDFFF, is ill-formed. The user can control how re2c treats such ill-formed sequences with the --encoding-policy <policy> switch.

For some encodings, there are code units that never occur in a valid encoded stream (e.g., 0xFF byte in UTF-8). If the generated scanner must check for invalid input, the only correct way to do so is to use the default rule (*). Note that the full range rule ([^]) won’t catch invalid code units when a variable-length encoding is used ([^] means “any valid code point”, whereas the default rule (*) means “any possible code unit”).

Start conditions

Conditions are enabled with -c --conditions. This option allows one to encode multiple interrelated lexers within the same re2c block.

Each lexer corresponds to a single condition. It starts with a label of the form yyc_name, where name is condition name and yyc prefix can be adjusted with configuration re2c:condprefix. Different lexers are separated with a comment /* *********************************** */ which can be adjusted with configuration re2c:cond:divider.

Furthermore, each condition has a unique identifier of the form yycname, where name is condition name and yyc prefix can be adjusted with configuration re2c:condenumprefix. Identifiers have the type YYCONDTYPE and should be generated with /*!types:re2c*/ directive or -t --type-header option. Users shouldn’t define these identifiers manually, as the order of conditions is not specified.

Before all conditions re2c generates entry code that checks the current condition identifier and transfers control flow to the start label of the active condition. After matching some rule of this condition, lexer may either transfer control flow back to the entry code (after executing the associated action and optionally setting another condition with =>), or use :=> shortcut and transition directly to the start label of another condition (skipping the action and the entry code). Configuration re2c:cond:goto allows one to change the default behavior.

Syntactically each rule must be preceded with a list of comma-separated condition names or a wildcard * enclosed in angle brackets < and >. Wildcard means “any condition” and is semantically equivalent to listing all condition names. Here regexp is a regular expression, default refers to the default rule *, and action is a block of code.

  • <conditions-or-wildcard>  regexp-or-default                 action

  • <conditions-or-wildcard>  regexp-or-default  =>  condition  action

  • <conditions-or-wildcard>  regexp-or-default  :=> condition

Rules with an exclamation mark ! in front of condition list have a special meaning: they have no regular expression, and the associated action is merged as an entry code to actions of normal rules. This might be a convenient place to peform a routine task that is common to all rules.

  • <!conditions-or-wildcard>  action

Another special form of rules with an empty condition list <> and no regular expression allows one to specify an “entry condition” that can be used to execute code before entering the lexer. It is semantically equivalent to a condition with number zero, name 0 and an empty regular expression.

  • <>                 action

  • <>  =>  condition  action

  • <>  :=> condition

Here is an example program that uses start conditions to parse integer numbers in binary, octal, decimal and hexadecimal format:

// re2c $INPUT -o $OUTPUT -ci
#include <stdint.h>
#include <limits.h>
#include <assert.h>

static const uint64_t ERROR = ~0lu;
/*!types:re2c*/

template<int BASE> static void adddgt(uint64_t &u, unsigned int d)
{
    u = u * BASE + d;
    if (u > UINT32_MAX) u = ERROR;
}

static uint64_t parse_u32(const char *s)
{
    const char *YYMARKER;
    int c = yycinit;
    uint64_t u = 0;

    /*!re2c
    re2c:yyfill:enable = 0;
    re2c:api:style = free-form;
    re2c:define:YYCTYPE = char;
    re2c:define:YYCURSOR = s;
    re2c:define:YYGETCONDITION = "c";
    re2c:define:YYSETCONDITION = "c = @@;";

    <*> * { return ERROR; }

    <init> '0b' / [01]        :=> bin
    <init> "0"                :=> oct
    <init> "" / [1-9]         :=> dec
    <init> '0x' / [0-9a-fA-F] :=> hex

    <bin, oct, dec, hex> "\x00" { return u; }

    <bin> [01]  { adddgt<2> (u, s[-1] - '0');      goto yyc_bin; }
    <oct> [0-7] { adddgt<8> (u, s[-1] - '0');      goto yyc_oct; }
    <dec> [0-9] { adddgt<10>(u, s[-1] - '0');      goto yyc_dec; }
    <hex> [0-9] { adddgt<16>(u, s[-1] - '0');      goto yyc_hex; }
    <hex> [a-f] { adddgt<16>(u, s[-1] - 'a' + 10); goto yyc_hex; }
    <hex> [A-F] { adddgt<16>(u, s[-1] - 'A' + 10); goto yyc_hex; }
    */
}

int main()
{
    assert(parse_u32("1234567890") == 1234567890);
    assert(parse_u32("0b1101") == 13);
    assert(parse_u32("0x7Fe") == 2046);
    assert(parse_u32("0644") == 420);
    assert(parse_u32("9999999999") == ERROR);
    assert(parse_u32("") == ERROR);
    return 0;
}

Skeleton programs

With the -S, --skeleton option, re2c ignores all non-re2c code and generates a self-contained C program that can be further compiled and executed. The program consists of lexer code and input data. For each constructed DFA (block or condition) re2c generates a standalone lexer and two files: an .input file with strings derived from the DFA and a .keys file with expected match results. The program runs each lexer on the corresponding .input file and compares results with the expectations. Skeleton programs are very useful for a number of reasons:

  • They can check correctness of various re2c optimizations (the data is generated early in the process, before any DFA transformations have taken place).

  • Generating a set of input data with good coverage may be useful for both testing and benchmarking.

  • Generating self-contained executable programs allows one to get minimized test cases (the original code may be large or have a lot of dependencies).

The difficulty with generating input data is that for all but the most trivial cases the number of possible input strings is too large (even if the string length is limited). Re2c solves this difficulty by generating sufficiently many strings to cover almost all DFA transitions. It uses the following algorithm. First, it constructs a skeleton of the DFA. For encodings with 1-byte code unit size (such as ASCII, UTF-8 and EBCDIC) skeleton is just an exact copy of the original DFA. For encodings with multibyte code units skeleton is a copy of DFA with certain transitions omitted: namely, re2c takes at most 256 code units for each disjoint continuous range that corresponds to a DFA transition. The chosen values are evenly distributed and include range bounds. Instead of trying to cover all possible paths in the skeleton (which is infeasible) re2c generates sufficiently many paths to cover all skeleton transitions, and thus trigger the corresponding conditional jumps in the lexer. The algorithm implementation is limited by ~1Gb of transitions and consumes constant amount of memory (re2c writes data to file as soon as it is generated).

Here is an example of a very simple program [example.re] that tries to match two-digit hexadecimal numbers:

/*!re2c
    *              {}
    [0-9a-fA-F]{2} {}
*/

We can see the generated DFA using `re2c -D example.re | dot -Grankdir=LR -Tpng -o example.png`:

../_images/example.png

Given this program, `re2c -S -o example.c example.re` generates three files: example.c (main program), example.c.line4.input (input data) and example.c.line4.keys (expected match results). First, let’s look at the generated strings [example.c.line4.input]:

$ hexdump -v -e '"%08_ax " 24/1 "%02x "' -e '" |" 24/1 "%_p" "|\n"' example.c.line4.input
00000000 00 01 02 03 04 05 06 07 08 09 0a 0b 0c 0d 0e 0f 10 11 12 13 14 15 16 17 |........................|
00000018 18 19 1a 1b 1c 1d 1e 1f 20 21 22 23 24 25 26 27 28 29 2a 2b 2c 2d 2e 2f |........ !"#$%&'()*+,-./|
00000030 3a 3b 3c 3d 3e 3f 40 47 48 49 4a 4b 4c 4d 4e 4f 50 51 52 53 54 55 56 57 |:;<=>?@GHIJKLMNOPQRSTUVW|
[ ... ]

Byte sequences correspond to the paths in DFA. All strings are glued together, so it’s hard to tell where is the end of one string and the beginning of another. For that re2c generates keys [example.c.line4.keys]:

$hexdump -v -e '"%08_ax " 36/1 "%02x " "\n"' example.c.line4.keys
00000000 01 01 fe 01 01 fe 01 01 fe 01 01 fe 01 01 fe 01 01 fe 01 01 fe 01 01 fe 01 01 fe 01 01 fe 01 01 fe 01 01 fe
00000024 01 01 fe 01 01 fe 01 01 fe 01 01 fe 01 01 fe 01 01 fe 01 01 fe 01 01 fe 01 01 fe 01 01 fe 01 01 fe 01 01 fe
00000048 01 01 fe 01 01 fe 01 01 fe 01 01 fe 01 01 fe 01 01 fe 01 01 fe 01 01 fe 01 01 fe 01 01 fe 01 01 fe 01 01 fe
[ ... ]

A key is a triplet: string length, the length of matching prefix and the number of matching rule. (If tags are used, there is an additional key per each tag variable). Keys are packed into an array. In our case each key fits into 1 byte, but in case of a larger DFA they might take 2 or 4 bytes. The generated program [example.c] contains two auxilary functions: read_file and action_line4. read_file is used to map .input and .keys files into memory (this function is shared between all lexers). action_line4 is a replacement for actions: it compares actual lexing results with the expected results. This function is specific to each lexer. Lexing is done by lex_line4: this function contains the generated DFA. The skeleton program is self-contained, so we can compile and run it:

$ cc -o example example.c
$ ./example
$ echo $?
0

When everything is fine (there are no errors) the program outputs nothing and exits with zero. For the sake of example, let’s pretend that re2c has an error that results in a missing case statement:

$ re2c -S -o example.c example.re
$ sed -i -e "s/case '7'://" example.c
$ cc -o example example.c
$ ./example
error: lex_line4: at position 248 (iteration 241):
        expected: match length 2, rule 0
        actual:   match length 1, rule 254

Viola! Now the lexer complains about discrepancy between expected and actual match results.

One special case of failure detected by skeleton programs is undefined control flow in the lexer. Use -Wundefined-control-flow warning to catch all such cases in in your code (more details here).

Visualization and debug

With the -D, --emit-dot option, re2c does not generate code. Instead, it dumps the generated DFA in DOT format. One can convert this dump to an image of the DFA using Graphviz or another library. Note that this option shows the final DFA after it has gone through a number of optimizations and transformations. Earlier stages can be dumped with various debug options, such as --dump-nfa, --dump-dfa-raw etc. (see the full list of options).

Below is an example of generating a picture of DFA that accepts any UTF-8 code point (utf8_any.re):

/*!re2c
    *   {}
    [^] {}
*/

Generate and render:

$ re2c -D -8 utf8_any.re | dot -Tpng -o utf8_any.png

Here is the picture:

../_images/utf8_any.png