How to Add Support for a New Language to Exuberant Ctags

Exuberant Ctags has been designed to make it very easy to add your own custom language parser. As an exercise, let us assume that I want to add support for my new language, Swine, the successor to Perl (i.e. Perl before Swine <wince>). This language consists of simple definitions of labels in the form "def my_label". Let us now examine the various ways to do this.

Operational background

As ctags considers each file name, it tries to determine the language of the file by applying the following three tests in order: if the file extension has been mapped to a language, if the file name matches a shell pattern mapped to a language, and finally if the file is executable and its first line specifies an interpreter using the Unix-style "#!" specification (if supported on the platform). If a language was identified, the file is opened and then the appropriate language parser is called to operate on the currently open file. The parser parses through the file and whenever it finds some interesting token, calls a function to define a tag entry.

Creating a user-defined language

The quickest and easiest way to do this is by defining a new language using the program options. In order to have Swine support available every time I start ctags, I will place the following lines into the file $HOME/.ctags, which is read in every time ctags starts:

  --langdef=swine
  --langmap=swine:.swn
  --regex-swine=/^def[ \t]*([a-zA-Z0-9_]+)/\1/d,definition/
The first line defines the new language, the second maps a file extension to it, and the third defines a regular expression to identify a language definition and generate a tag file entry for it.

Integrating a new language parser

Now suppose that I want to truly integrate compiled-in support for Swine into ctags. First, I create a new module, swine.c, and add one externally visible function to it, extern parserDefinition *SwineParser(void), and add its name to the table in parsers.h. The job of this parser definition function is to create an instance of the parserDefinition structure (using parserNew()) and populate it with information defining how files of this language are recognized, what kinds of tags it can locate, and the function used to invoke the parser on the currently open file.

The structure parserDefinition allows assignment of the following fields:

  const char *name;               /* name of language */
  kindOption *kinds;              /* tag kinds handled by parser */
  unsigned int kindCount;         /* size of `kinds' list */
  const char *const *extensions;  /* list of default extensions */
  const char *const *patterns;    /* list of default file name patterns */
  parserInitialize initialize;    /* initialization routine, if needed */
  simpleParser parser;            /* simple parser (common case) */
  rescanParser parser2;           /* rescanning parser (unusual case) */
  boolean regex;                  /* is this a regex parser? */

The name field must be set to a non-empty string. Also, unless regex is set true (see below), either parser or parser2 must set to point to a parsing routine which will generate the tag entries. All other fields are optional.

Now all that is left is to implement the parser. In order to do its job, the parser should read the file stream using using one of the two I/O interfaces: either the character-oriented fileGetc(), or the line-oriented fileReadLine(). When using fileGetc(), the parser can put back a character using fileUngetc(). How our Swine parser actually parses the contents of the file is entirely up to the writer of the parser--it can be as crude or elegant as desired. You will note a variety of examples from the most complex (c.c) to the simplest (make.c).

When the Swine parser identifies an interesting token for which it wants to add a tag to the tag file, it should create a tagEntryInfo structure and initialize it by calling initTagEntry(), which initializes defaults and fills information about the current line number and the file position of the beginning of the line. After filling in information defining the current entry (and possibly overriding the file position or other defaults), the parser passes this structure to makeTagEntry().

Instead of writing a character-oriented parser, it may be possible to specify regular expressions which define the tags. In this case, instead of defining a parsing function, SwineParser(), sets regex to true, and points initialize to a function which calls addTagRegex() to install the regular expressions which define its tags. The regular expressions thus installed are compared against each line of the input file and generate a specified tag when matched. It is usually much easier to write a regex-based parser, although they can be slower (one parser example was 4 times slower). Whether the speed difference matters to you depends upon how much code you have to parse. It is probably a good strategy to implement a regex-based parser first, and if it is too slow for you, then invest the time and effort to write a character-based parser.

A regex-based parser is inherently line-oriented (i.e. the entire tag must be recognizable from looking at a single line) and context-insensitive (i.e the generation of the tag is entirely based upon when the regular expression matches a single line). However, a regex-based callback mechanism is also available, installed via the function addCallbackRegex(). This allows a specified function to be invoked whenever a specific regular expression is matched. This allows a character-oriented parser to operate based upon context of what happened on a previous line (e.g. the start or end of a multi-line comment). Note that regex callbacks are called just before the first character of that line can is read via either fileGetc() or using fileGetc(). The effect of this is that before either of these routines return, a callback routine may be invoked because the line matched a regex callback. A callback function to be installed is defined by these types:

  typedef void (*regexCallback) (const char *line, const regexMatch *matches, unsigned int count);

  typedef struct {
      size_t start;   /* character index in line where match starts */
      size_t length;  /* length of match */
  } regexMatch;

The callback function is passed the line matching the regular expression and an array of count structures defining the subexpression matches of the regular expression, starting from \0 (the entire line).

Lastly, be sure to add your the name of the file containing your parser (e.g. swine.c) to the macro SOURCES in the file source.mak and an entry for the object file to the macro OBJECTS in the same file, so that your new module will be compiled into the program.

This is all there is to it. All other details are specific to the parser and how it wants to do its job. There are some support functions which can take care of some commonly needed parsing tasks, such as keyword table lookups (see keyword.c), which you can make use of if desired (examples of its use can be found in c.c, eiffel.c, and fortran.c). Almost everything is already taken care of automatically for you by the infrastructure. Writing the actual parsing algorithm is the hardest part, but is not constrained by any need to conform to anything in ctags other than that mentioned above.

There are several different approaches used in the parsers inside Exuberant Ctags and you can browse through these as examples of how to go about creating your own.

Examples

Below you will find several example parsers demonstrating most of the facilities available. These include three alternative implementations of a Swine parser, which generate tags for lines beginning with "def" followed by some name.

/***************************************************************************
 * swine.c
 * Character-based parser for Swine definitions
 **************************************************************************/
/* INCLUDE FILES */
#include "general.h"    /* always include first */

#include <string.h>     /* to declare strxxx() functions */
#include <ctype.h>      /* to define isxxx() macros */

#include "parse.h"      /* always include */
#include "read.h"       /* to define file fileReadLine() */

/* DATA DEFINITIONS */
typedef enum eSwineKinds {
    K_DEFINE
} swineKind;

static kindOption SwineKinds [] = {
    { TRUE, 'd', "definition", "pig definition" }
};

/* FUNCTION DEFINITIONS */

static void findSwineTags (void)
{
    vString *name = vStringNew ();
    const unsigned char *line;

    while ((line = fileReadLine ()) != NULL)
    {
        /* Look for a line beginning with "def" followed by name */
        if (strncmp ((const char*) line, "def", (size_t) 3) == 0  &&
            isspace ((int) line [3]))
        {
            const unsigned char *cp = line + 4;
            while (isspace ((int) *cp))
                ++cp;
            while (isalnum ((int) *cp)  ||  *cp == '_')
            {
                vStringPut (name, (int) *cp);
                ++cp;
            }
            vStringTerminate (name);
            makeSimpleTag (name, SwineKinds, K_DEFINE);
            vStringClear (name);
        }
    }
    vStringDelete (name);
}

/* Create parser definition stucture */
extern parserDefinition* SwineParser (void)
{
    static const char *const extensions [] = { "swn", NULL };
    parserDefinition* def = parserNew ("Swine");
    def->kinds      = SwineKinds;
    def->kindCount  = KIND_COUNT (SwineKinds);
    def->extensions = extensions;
    def->parser     = findSwineTags;
    return def;
}


/***************************************************************************
 * swine.c
 * Regex-based parser for Swine
 **************************************************************************/
/* INCLUDE FILES */
#include "general.h"    /* always include first */
#include "parse.h"      /* always include */

/* FUNCTION DEFINITIONS */

static void installSwineRegex (const langType language)
{
    addTagRegex (language, "^def[ \t]*([a-zA-Z0-9_]+)", "\\1", "d,definition", NULL);
}

/* Create parser definition stucture */
extern parserDefinition* SwineParser (void)
{
    static const char *const extensions [] = { "swn", NULL };
    parserDefinition* def = parserNew ("Swine");
    parserDefinition* const def = parserNew ("Makefile");
    def->patterns   = patterns;
    def->extensions = extensions;
    def->initialize = installMakefileRegex;
    def->regex      = TRUE;
    return def;
}

/***************************************************************************
 * swine.c
 * Regex callback-based parser for Swine definitions
 **************************************************************************/
/* INCLUDE FILES */
#include "general.h"    /* always include first */

#include "parse.h"      /* always include */
#include "read.h"       /* to define file fileReadLine() */

/* DATA DEFINITIONS */
typedef enum eSwineKinds {
    K_DEFINE
} swineKind;

static kindOption SwineKinds [] = {
    { TRUE, 'd', "definition", "pig definition" }
};

/* FUNCTION DEFINITIONS */

static void definition (const char *const line, const regexMatch *const matches,
                       const unsigned int count)
{
    if (count > 1)    /* should always be true per regex */
    {
        vString *const name = vStringNew ();
        vStringNCopyS (name, line + matches [1].start, matches [1].length);
        makeSimpleTag (name, SwineKinds, K_DEFINE);
    }
}

static void findSwineTags (void)
{
    while (fileReadLine () != NULL)
        ;  /* don't need to do anything here since callback is sufficient */
}

static void installSwine (const langType language)
{
    addCallbackRegex (language, "^def[ \t]+([a-zA-Z0-9_]+)", NULL, definition);
}

/* Create parser definition stucture */
extern parserDefinition* SwineParser (void)
{
    static const char *const extensions [] = { "swn", NULL };
    parserDefinition* def = parserNew ("Swine");
    def->kinds      = SwineKinds;
    def->kindCount  = KIND_COUNT (SwineKinds);
    def->extensions = extensions;
    def->parser     = findSwineTags;
    def->initialize = installSwine;
    return def;
}

/***************************************************************************
 * make.c
 * Regex-based parser for makefile macros
 **************************************************************************/
/* INCLUDE FILES */
#include "general.h"    /* always include first */
#include "parse.h"      /* always include */

/* FUNCTION DEFINITIONS */

static void installMakefileRegex (const langType language)
{
    addTagRegex (language, "(^|[ \t])([A-Z0-9_]+)[ \t]*:?=", "\\2", "m,macro", "i");
}

/* Create parser definition stucture */
extern parserDefinition* MakefileParser (void)
{
    static const char *const patterns [] = { "[Mm]akefile", NULL };
    static const char *const extensions [] = { "mak", NULL };
    parserDefinition* const def = parserNew ("Makefile");
    def->patterns   = patterns;
    def->extensions = extensions;
    def->initialize = installMakefileRegex;
    def->regex      = TRUE;
    return def;
}