C Left-Right Parser (libcleri)

Language parser for the C/C++ programming language. Initially created for SiriDB.

---

• Installation
• Related projects
• Quick usage
• API
  • cleri_t
  • cleri_grammar_t
  • cleri_parse_t
  • cleri_node_t
  • cleri_children_t
  • cleri_olist_t
• Elements
  • cleri_keyword_t
  • cleri_regex_t
  • cleri_choice_t
  • cleri_sequence_t
  • cleri_optional_t
  • cleri_prio_t
  • cleri_repeat_t
  • cleri_list_t
  • cleri_token_t
  • cleri_tokens_t
  • Forward reference
  • cleri_dup_t
• Miscellaneous functions

---

Installation

Note: libcleri requires pcre2

On Ubuntu:

sudo apt install libpcre2-dev

On MacOs:

brew install pcre2

Install debug or release version, in this example we will install the release version.

$ cd Release

Compile libcleri

Note: On MacOs you might need to set environment variables:

export CFLAGS="-I/usr/local/include" && export LDFLAGS="-L/usr/local/lib"

$ make all

Install libcleri

$ sudo make install

Note: run sudo make uninstall for removal.

Related projects

• pyleri: Python parser (can export grammar to pyleri, libcleri, goleri and jsleri)
• jsleri: JavaScript parser
• goleri: Go parser

Quick usage

The recommended way to create a grammar is to use pyleri for writing the grammar and then export the grammar to libcleri or other languages.

This is a simple example using libcleri:

#include <stdio.h>
#include <cleri/cleri.h>

void test_str(cleri_grammar_t * grammar, const char * str)
{
    cleri_parse_t * pr = cleri_parse(grammar, str);
    printf("Test string '%s': %s\n", str, pr->is_valid ? "true" : "false");
    cleri_parse_free(pr);
}

int main(void)
{
    /* define grammar */
    cleri_t * k_hi = cleri_keyword(0, "hi", 0);
    cleri_t * r_name = cleri_regex(0, "^(?:\"(?:[^\"]*)\")+");
    cleri_t * start = cleri_sequence(0, 2, k_hi, r_name);

    /* compile grammar */
    cleri_grammar_t * my_grammar = cleri_grammar(start, NULL);

    /* test some strings */
    test_str(my_grammar, "hi \"Iris\"");  // true
    test_str(my_grammar, "bye \"Iris\""); // false

    /* cleanup grammar */
    cleri_grammar_free(my_grammar);

    return 0;
}

Although libcleri is written for C, it can be used with C++ too:

#include <iostream>
#include <cleri/cleri.h>

void test_str(cleri_grammar_t * grammar, const char * str)
{
    cleri_parse_t * pr = cleri_parse(grammar, str);
    std::cout << "Test string " << str << ": " <<
            (pr->is_valid ? "true" : "false") << std::endl;
    cleri_parse_free(pr);
}

int main()
{
    /* define grammar */
    cleri_t * k_hi = cleri_keyword(0, "hi", 0);
    cleri_t * r_name = cleri_regex(0, "^(?:\"(?:[^\"]*)\")+");
    cleri_t * start = cleri_sequence(0, 2, k_hi, r_name);

    /* compile grammar */
    cleri_grammar_t * my_grammar = cleri_grammar(start, NULL);

    /* test some strings */
    test_str(my_grammar, "hi \"Iris\"");  // true
    test_str(my_grammar, "bye \"Iris\""); // false

    /* cleanup grammar */
    cleri_grammar_free(my_grammar);

    return 0;
}

API

cleri_t

Cleri type is the base object for each element.

Public members

• uint32_t gid: Global Identifier for the element. This GID is not required and as a rule it should be set to 0 if not used. You can use the GID for identifiying an element in a parse result. When exporting a Pyleri grammar, each named element automatically gets a unique GID assigned. (readonly)
• cleri_tp tp: Type for the cleri object. (readonly)
  • CLERI_TP_SEQUENCE
  • CLERI_TP_OPTIONAL
  • CLERI_TP_CHOICE
  • CLERI_TP_LIST
  • CLERI_TP_REPEAT
  • CLERI_TP_PRIO
  • CLERI_TP_RULE
  • CLERI_TP_THIS
  • CLERI_TP_KEYWORD
  • CLERI_TP_TOKEN
  • CLERI_TP_TOKENS
  • CLERI_TP_REGEX
  • CLERI_TP_END_OF_STATEMENT
• cleri_via_t via: Element. (readonly)
  • cleri_sequence_t * sequence
  • cleri_optional_t * optional
  • cleri_choice_t * choice
  • cleri_list_t * list
  • cleri_repeat_t * repeat
  • cleri_prio_t * prio
  • cleri_rule_t * rule
  • cleri_keyword_t * keyword
  • cleri_regex_t * regex
  • cleri_token_t * token
  • cleri_tokens_t * tokens
  • void * dummy (place holder, this, eof)

cleri_t * cleri_new(uint32_t gid, cleri_tp tp, cleri_free_object_t free_object, cleri_parse_object_t parse_object)

Create and return a new cleri object. A unique gid is not required but can help you with identifiying the element in a parse result. As a rule you should assign 0 in case no specific gid is required. This function should only be used in case you want to create your own custom element.

void cleri_incref(cleri_t * cl_object)

Increment the reference counter for a cleri object. Should only be used in case you want to write your own custom element.

void cleri_decref(cleri_t * cl_object)

Decrement the reference counter for a cleri object. If no references are left the object will be destroyed. Do not use this function after the element has successfully been added to another element or grammar. Should only be used in case you want to write your own custom element.

int cleri_free(cleri_t * cl_object)

Decrement reference counter for a cleri object. When there are no more references left the object will be destroyed. Use this function to cleanup after errors have occurred. Do not use this function after the element has successfully been added to another element or grammar.

Example strict error handling:

cleri_grammar_t * compile_grammar(void)
{
    cleri_t * k_hello = cleri_keyword(0, "hello", 0);
    if (k_hello == NULL) {
        return NULL;
    }
    cleri_t * k_world = cleri_keyword(0, "world", 0);
    if (k_world == NULL) {
        cleri_free(k_hello); // must cleanup k_hello
        return NULL;
    }
    cleri_t * hello_world = cleri_sequence(0, 2, k_hello, k_world);
    if (start == NULL) {
        cleri_free(k_hello);
        cleri_free(k_world);
        return NULL;
    }
    cleri_t * opt = cleri_optional(0, hello_world);
    if (opt == NULL) {
        /* we now must only cleanup hello_world since this sequence will
        * cleanup both keywords too. */
        cleri_free(hello_world);
        return NULL;
    }
    cleri_grammar_t * grammar = cleri_grammar(opt, NULL);
    if (grammar == NULL) {
        cleri_free(opt);
    }
    /* when your program has finished, the grammar including all elements can
     * be destroyed using cleri_grammar_free() */
    return grammar;
}

Note: Usually grammar is only compiled at the startup of your program so memory allocation errors during the grammar creation are unlikely to occur. If NULL is parsed as an argument instead of an element, then the function to which the argument is parsed to, will return NULL. Following this chain the final grammar returns NULL in case an error has occurred somewhere. In this case you should usually abort the program.

cleri_grammar_t

Compiled libcleri grammar.

No public members

cleri_grammar_t * cleri_grammar(cleri_t * start, const char * re_keywords)

Create and return a compiled grammar. Argument start must be the entry element for the grammar. Argument re_keywords should be a regular expression starting with character ^ for matching keywords in a grammar. When a grammar is created, each defined keyword should match this regular expression. re_keywords is allowed to be NULL in which case the defualt CLERI_DEFAULT_RE_KEYWORDS is used.

void cleri_grammar_free(cleri_grammar_t * grammar)

Cleanup grammar. This will also destroy all elements which are used by the grammar. Make sure all parse results are destroyed before destroying the grammar because a parse result depends on elements from the grammar.

cleri_parse_t

Parse result containing the parse tree and other information about the parse result.

Public members

• int cleri_parse_t.is_valid: Boolean. Value is 1 (TRUE) in case the parse string is valid or 0 (FALSE) if not. (readonly)
• size_t cleri_parse_t.pos: Position in the string to where the string was successfully parsed. This value is (readonly) equal to the length of the string in case cleri_parse_t.is_valid is TRUE. (readonly)
• const char * cleri_parse_t.str: Pointer to the provided string. (readonly)
• cleri_node_t * tree: Parse tree. (see cleri_node_t and cleri_children_t) (readonly)
• const cleri_olist_t * expect: Linked list to possible elements at position cleri_parse_t.pos in cleri_parse_t.str. (see cleri_olist_t for more information)

cleri_parse_t * cleri_parse(cleri_grammar_t * grammar, const char * str)

Create and return a parse result. The parse result contains pointers to the provided string (str) so make sure the string is available while using the parse result.

void cleri_parse_free(cleri_parse_t * pr)

Cleanup a parse result.

void cleri_parse_expect_start(cleri_parse_t * pr)

Can be used to reset the expect list to start. Usually you are not required to use this function since the expect list is already at the start position.

cleri_node_t

Node object. A parse result has a parse tree which consists of nodes. Each node may have children.

Public members

• const char * cleri_node_t.str: Pointer to the position in the parse string where this node starts. (readonly)
• size_t cleri_node_t.len: Length of the string which is applicable for this node. (readonly)
• cleri_t * cleri_node_t.cl_obj: Element from the grammar which matches this node. (readonly)
• cleri_children_t * cleri_node_t.children: Optional children for this node. (readonly)

bool cleri_node_has_children(cleri_node_t * node)

Macro function for checking if a node has children.

cleri_children_t

Children from a node in a linked list.

Public members

• cleri_node_t * cleri_children_t.node: Child node. (readonly)
• struct cleri_children_s * cleri_children_t.next: Next child node or NULL if there are no other childs. (readonly)

Example looping over all children within a node:

/* we asume having a node (cleri_node_t*) */
if (cleri_node_has_children(node)) {
    cleri_children_t * child = node->children;
    while (child != NULL) {
        // do something with child->node
        child = child->next;
    }
}

cleri_olist_t

Linked list holding libcleri objects. A cleri_olist_t type is used for expected elements in a parse result.

Public members

• cleri_t * cl_obj: Object (holding an element, readonly)
• cleri_olist_t * next: Next object. (readonly)

Example looping over cleri_parse_t.expect:

/* we assume having a pr (cleri_parse_t*)
 *
 * Notes:
 *    pr->expect is NULL if nothing is expected and it is save to
 *    change pr->expect. If required the linked list can be reset to start
 *    using cleri_parse_expect_start(). */
while (pr->expect != NULL) {
    // do something with pr->expect->cl_obj
    pr->expect = pr->expect->next;
}

Elements

Elements are objects used to define a grammar.

cleri_keyword_t

Keyword element. The parser needs a match with the keyword.

Type (cleri_t.tp): CLERI_TP_KEYWORD

Public members

• const char * cleri_keyword_t.keyword: Contains the keyword string. (readonly)
• int cleri_keyword_t.ign_case: Boolean. (readonly)
• size_t cleri_keyword_t.len: Length of the keyword string. (readonly)

cleri_t * cleri_keyword(uint32_t gid, const char * keyword, int ign_case)

Create and return a new object containing a keyword element. Argument ign_case can be set to 1 for a case insensitive keyword match.

Example:

/* define case insensitive keyword */
cleri_t * k_tictactoe = cleri_keyword(
    0,                  // gid, not used in this example
    "tic-tac-toe",      // keyword
    1);                 // case insensitive

/* create grammar with custom keyword regular expression match */
cleri_grammar_t * grammar = cleri_grammar(k_tictactoe, "^[A-Za-z-]+");

/* parse some test string */
cleri_parse_t * pr = cleri_parse(grammar, "Tic-Tac-Toe");
printf("Valid: %s\n", pr->is_valid ? "true" : "false"); // true

/* cleanup */
cleri_parse_free(pr);
cleri_grammar_free(grammar);

cleri_regex_t

Regular expression element. The parser needs a match with the regular expression.

No public members

cleri_t * cleri_regex(uint32_t gid, const char * pattern)

Create and return a new object containing a regular expression element. Argument pattern should contain the regular expression. Each pattern must start with character ^ and the pattern should be checked before calling this function.

See Quick usage for a cleri_regex_t example.

cleri_choice_t

Choice element. The parser must choose one of the child elements.

Public members

• int cleri_choice_t.most_greedy: Boolean. (readonly)
• cleri_olist_t * cleri_choice_t.olist: Children. (readonly)

cleri_t * cleri_choice(uint32_t gid, int most_greedy, size_t len, ...)

Create and return a new object containing a choice element. Argument most_greedy can be set to 1 in which case the parser will select the most greedy match. When 0, the parser will select the first match.

Example:

/* define grammar */
cleri_t * k_hello = cleri_keyword(0, "hello", 0);
cleri_t * k_goodbye = cleri_keyword(0, "goodbye", 0);
cleri_t * choice = cleri_choice(
    0,                      // gid, not used in this example
    0,                      // stop at first match
    2,                      // number of elements
    k_hello, k_goodbye);    // elements

/* create grammar */
cleri_grammar_t * grammar = cleri_grammar(choice, NULL);

/* parse some test string */
cleri_parse_t * pr = cleri_parse(grammar, "goodbye");
printf("Valid: %s\n", pr->is_valid ? "true" : "false"); // true

/* cleanup */
cleri_parse_free(pr);
cleri_grammar_free(grammar);

cleri_sequence_t

Sequence element. The parser must match each element in the specified order.

Public members

• cleri_olist_t * cleri_sequence_t.olist: Elements. (readonly)

cleri_t * cleri_sequence(uint32_t gid, size_t len, ...)

Create and return a new object containing a sequence element.

Example:

cleri_t * sequence = cleri_sequence(
    0,                              // gid, not used in the example
    3,                              // number of elements
    cleri_keyword(0, "Tic", 0),     // first element
    cleri_keyword(0, "Tac", 0),     // second element
    cleri_keyword(0, "Toe", 0));    // third element

/* create grammar */
cleri_grammar_t * grammar = cleri_grammar(sequence, NULL);

/* parse some test string */
cleri_parse_t * pr = cleri_parse(grammar, "Tic Tac Toe");
printf("Valid: %s\n", pr->is_valid ? "true" : "false"); // true

/* cleanup */
cleri_parse_free(pr);
cleri_grammar_free(grammar);

cleri_optional_t

Optional element. The parser looks for an optional element.

Public members

• cleri_t * cleri_optional_t.cl_obj: Optional element. (readonly)

cleri_t * cleri_optional(uint32_t gid, cleri_t * cl_obj)

Create and return a new object containing an optional element.

Example:

/* define grammar */
cleri_t * k_hello = cleri_keyword(0, "hello", 0);
cleri_t * k_there = cleri_keyword(0, "there", 0);
cleri_t * optional = cleri_optional(
    0,                  // gid, not used in this example
    k_there);           // optional element
cleri_t * greet = cleri_sequence(
    0,                  // gid, not used in this example
    2,                  // number of elements
    k_hello, optional); // elements

/* create grammar */
cleri_grammar_t * grammar = cleri_grammar(greet, NULL);

/* parse some test string */
cleri_parse_t * pr = cleri_parse(grammar, "hello");
printf("Valid: %s\n", pr->is_valid ? "true" : "false"); // true

/* cleanup */
cleri_parse_free(pr);
cleri_grammar_free(grammar);

cleri_prio_t

Prio element. The parser must match one element. Inside the prio element it is possible to use CLERI_THIS which is a reference to itself.

Note: Use a forward reference when possible. A prio is required when the same position in a string is potentially checked more than once.

Public members

• cleri_olist_t * cleri_sequence_t.olist: Elements. (readonly)

cleri_t * cleri_prio(uint32_t gid, size_t len, ...)

Create and return a new object containing a prio element.

Example:

/*
 * define grammar.
 *
 * Note: The third and fourth element are using a reference to the prio
 *       element at the same position in the string as the prio element.
 *       This is why a forward reference cannot be used for this example.
 */
cleri_t * prio = cleri_prio(
    0,                              // gid, not used in the example
    4,                              // number of elements
    cleri_keyword(0, "ni", 0),      // first element
    cleri_sequence(0, 3,            // second element
        cleri_token(0, "("),
        CLERI_THIS,
        cleri_token(0, ")")),
    cleri_sequence(0, 3,            // third element
        CLERI_THIS,
        cleri_keyword(0, "or", 0),
        CLERI_THIS),
    cleri_sequence(0, 3,            // fourth element
        CLERI_THIS,
        cleri_keyword(0, "and", 0),
        CLERI_THIS));

/* create grammar */
cleri_grammar_t * grammar = cleri_grammar(prio, NULL);

/* parse some test string */
cleri_parse_t * pr = cleri_parse(grammar, "(ni or ni) and (ni or ni)");
printf("Valid: %s\n", pr->is_valid ? "true" : "false"); // true

/* cleanup */
cleri_parse_free(pr);
cleri_grammar_free(grammar);

cleri_repeat_t

Repeat element. The parser must math at least cleri_repeat_t.min elements and at most cleri_repeat_t.max. An unlimited amount is allowed in case cleri_repeat_t.max is set to 0 (zero).

Public members

• cleri_t * cleri_repeat_t.cl_obj: Element to repeat. (readonly)
• size_t cleri_repeat_t.min: Minimum times an element is expected. (readonly)
• size_t cleri_repeat_t.max: Maximum times an element is expected or 0 for unlimited. (readonly)

cleri_t * cleri_repeat(uint32_t gid, cleri_t * cl_obj, size_t min, size_t max)

Create and return a new object containing a repeat element. Argument max should be greater or equal to min or 0.

Example:

/* define grammar */
cleri_t * repeat = cleri_repeat(
    0,                          // gid, not used in this example
    cleri_keyword(0, "ni", 0),  // repeated element
    0,                          // min n times
    0);                         // max n times (0 for unlimited)

/* create grammar */
cleri_grammar_t * grammar = cleri_grammar(repeat, NULL);

/* parse some test string */
cleri_parse_t * pr = cleri_parse(grammar, "ni ni ni ni ni");
printf("Valid: %s\n", pr->is_valid ? "true" : "false"); // true

/* cleanup */
cleri_parse_free(pr);
cleri_grammar_free(grammar);

cleri_list_t

List element. Like repeat but with a delimiter.

Public members

• cleri_t * cleri_list_t.cl_obj: Element to repeat. (readonly)
• cleri_t * cleri_list_t.delimiter: Delimiter between repeating element. (readonly)
• size_t cleri_list_t.min: Minimum times an element is expected. (readonly)
• size_t cleri_list_t.max: Maximum times an element is expected or 0 for unlimited. (readonly)
• int cleri_list_t.opt_closing: Allow or disallow ending with a delimiter.

cleri_t * cleri_list(uint32_t gid, cleri_t * cl_obj, cleri_t * delimiter, size_t min, size_t max, int opt_closing)

Create and return a new object containing a list element. Argument max should be greater or equal to min or 0. Argument opt_closing can be 1 (TRUE) to allow or 0 (FALSE) to disallow a list to end with a delimiter.

Example:

/* define grammar */
cleri_t * list = cleri_list(
    0,                          // gid, not used in this example
    cleri_keyword(0, "ni", 0),  // repeated element
    cleri_token(0, ","),        // delimiter element
    0,                          // min n times
    0,                          // max n times (0 for unlimited)
    0);                         // disallow ending with a delimiter

/* create grammar */
cleri_grammar_t * grammar = cleri_grammar(list, NULL);

/* parse some test string */
cleri_parse_t * pr = cleri_parse(grammar, "ni, ni, ni, ni, ni");
printf("Valid: %s\n", pr->is_valid ? "true" : "false"); // true

/* cleanup */
cleri_parse_free(pr);
cleri_grammar_free(grammar);

cleri_token_t

Token element. The parser must math a token exactly. A token can be one or more characters and is usually used to match operators like +, -, * etc.

Public members

• const char * cleri_token_t.token: Token string. (readonly)
• size_t cleri_token_t.len: Length of the token string. (readonly)

cleri_t * cleri_token(uint32_t gid, const char * token)

Create and return a new object containing a token element.

Example:

/* define grammar */
cleri_t * token = cleri_token(
    0,          // gid, not used in this example
    "-");       // token string (dash)

cleri_t * ni =  cleri_keyword(0, "ni", 0);
cleri_t * list = cleri_list(0, ni, token, 0, 0, 0);

/* create grammar */
cleri_grammar_t * grammar = cleri_grammar(list, NULL);

/* parse some test string */
cleri_parse_t * pr = cleri_parse(grammar, "ni-ni - ni- ni -ni");
printf("Valid: %s\n", pr->is_valid ? "true" : "false"); // true

/* cleanup */
cleri_parse_free(pr);
cleri_grammar_free(grammar);

cleri_tokens_t

Tokens element. Can be used to register multiple tokens at once.

cleri_t * cleri_tokens(uint32_t gid, const char * tokens)

Create and return a new object containing a tokens element. Argument tokens must be a string with tokens seperated by spaces. If given tokens are different in size, the parser will try to match the longest tokens first.

Example:

/* define grammar */
cleri_t * tokens = cleri_tokens(
    0,              // gid, not used in this example
    "+ - -=");      // tokens string '+', '-' and '-='

cleri_t * ni =  cleri_keyword(0, "ni", 0);
cleri_t * list = cleri_list(0, ni, tokens, 0, 0, 0);

/* create grammar */
cleri_grammar_t * grammar = cleri_grammar(list, NULL);

/* parse some test string */
cleri_parse_t * pr = cleri_parse(grammar, "ni + ni -= ni - ni");
printf("Valid: %s\n", pr->is_valid ? "true" : "false"); // true

/* cleanup */
cleri_parse_free(pr);
cleri_grammar_free(grammar);

Forward reference

Forward reference to a libcleri object. There is no specific type for a reference.

Warning: A reference is not protected against testing the same position in in a string. This could potentially lead to an infinite loop. For example:

cleri_ref_set(ref, cleri_optional(0, ref)); // DON'T DO THIS

Use prio if such recursive construction is required.

cleri_t * cleri_ref(void)

Create and return a new object as reference element. Once the reference is created, it can be used as element in you grammar. Do not forget to actually set the reference using cleri_ref_set().

void cleri_ref_set(cleri_t * ref, cleri_t * cl_obj)

Set a reference. For every created forward reference, this function must be called exactly once. Argument ref must be created with cleri_ref(). Argument cl_obj cannot be used outside the reference. Since the reference becomes the cl_obj, it is the reference you should use.

Example

/* define grammar */
cleri_t * ref = cleri_ref();
cleri_t * choice = cleri_choice(
    0, 0, 2, cleri_keyword(0, "ni", 0), ref);

cleri_ref_set(ref, cleri_sequence(
    0,
    3,
    cleri_token(0, "["),
    cleri_list(0, choice, cleri_token(0, ","), 0, 0, 0),
    cleri_token(0, "]")));

/* create grammar */
cleri_grammar_t * grammar = cleri_grammar(ref, NULL);

/* parse some test string */
cleri_parse_t * pr = cleri_parse(grammar, "[ni, ni, [ni, [], [ni, ni]]]");
printf("Valid: %s\n", pr->is_valid ? "true" : "false"); // true

/* cleanup */
cleri_parse_free(pr);
cleri_grammar_free(grammar);

cleri_dup_t

Duplicate an object. The type is an extension to cleri_t.

cleri_t * cleri_dup(uint32_t gid, cleri_t * cl_obj)

Duplicate a libcleri object with a different gid but using the same element.

Note: Only the object is duplicated. The element (cleri_via_t via) is a pointer to the original object.

The following pyleri code will use cleri_dup() when exported to c:

elem = Repeat(obj, mi=1, ma=1)

Use the code below if you want similar behavior without duplication:

elem = Sequence(obj)

Miscellaneous functions

const char * cleri_version(void)

Returns the version of libcleri.