lsup_rdf/include/term.h

#ifndef _LSUP_TERM_H
#define _LSUP_TERM_H

#include <assert.h>
#include <regex.h>

#include "buffer.h"
#include "namespace.h"

#define UUID4_URN_SIZE UUIDSTR_SIZE + 10

// Some common RDF term values.
#define LSUP_RDF_TYPE       "http://www.w3.org/1999/02/22-rdf-syntax-ns#type"
#define LSUP_RDF_TYPE_NS    "rdf:type"
/// Default data type for untyped literals (prefixed IRI).
#define DEFAULT_DTYPE       "http://www.w3.org/2001/XMLSchema#string"
#define DEFAULT_DTYPE_NS    "xsd:string"

/** @brief URI parsing regular expression.
 *
 * Based on RFC3986 (see https://tools.ietf.org/html/rfc3986#appendix-B) and
 * modified for use in this application. Relevant matching groups are the
 * following, for a sample URI `http://example.org/123/456/?query=blah#frag`:
 *
 * #0:  Full parsed URI (http://example.org/123/456/?query=blah#frag)
 * #1:  Domain prefix (http://example.org)
 * #2:  Protocol (http:)
 * #4:  Authority (example.org)
 * #5:  Path relative to domain (/123/456/?query=blah#frag)
 * #6:  Path, excluding query and fragment (/123/456/)
 * #8:  Query (query=blah)
 * #10: Fragment (frag)
 *
 * For URN-like URIs, such as `urn:s:0`, the prefix part (#1) is `urn:` and
 * the path (#4) is `s:0`.
 */
#define LSUP_URI_REGEX_STR \
    "^(([^:/?#]+:)?(//([^/?#]*))?)?(([^?#]*)(\\?([^#]*))?(#(.*))?)"


/*
 * Data types.
 */

/// Language tag, currently restricted to 7 characters.
typedef char LSUP_LangTag[8];

/// Term type.
typedef enum {
    LSUP_TERM_UNDEFINED = 0,/**<
                              * Undefined placeholder or result of an error.
                              * Invalid for most operations.
                              */
    LSUP_TERM_IRIREF,       ///< IRI reference.
    LSUP_TERM_NS_IRIREF,    ///< Namespace-prefixed IRI reference.
    LSUP_TERM_LITERAL,      ///< Literal without language tag.
    LSUP_TERM_LT_LITERAL,   ///< Language-tagged string literal.
    LSUP_TERM_BNODE,        ///< Blank node.
} LSUP_TermType;

/** @brief IRI information.
 *
 * See regex matching group for #LSUP_URI_REGEX_STR for more information.
 */
typedef struct iri_info_t LSUP_IRIInfo;

typedef struct link_map_iter LSUP_LinkMapIterator;

/// RDF term.
typedef struct term_t {
    char *              data;       // URI, literal value, or BNode label.
    union {
        struct term_t * datatype;   // Data type IRI for LSUP_TERM_LITERAL.
        LSUP_LangTag    lang;       // Lang tag for LSUP_TERM_LT_LITERAL.
        LSUP_Key        bnode_id;   // BNode ID for comparison & skolemization.
        LSUP_IRIInfo *  iri_info;   // IRI information structure.
    };
    LSUP_TermType       type;       // Term type.
} LSUP_Term;


/** @brief Shorthand to test if a term is a IRI of any kind.
 */
#define LSUP_IS_IRI(term) \
    ((term)->type == LSUP_TERM_IRIREF || (term)->type == LSUP_TERM_NS_IRIREF)

/** @brief Shorthand to test if a term is a literal of any kind.
 */
#define LSUP_IS_LITERAL(term) \
    ((term)->type == LSUP_TERM_LITERAL || (term)->type == LSUP_TERM_LT_LITERAL)


/** @brief RDF triple.
 *
 * This represents a complete RDF statement. Triple terms can be accessed
 * directly via the `s`, `p`, `o` members or sequentially via
 * #LSUP_triple_pos().
 */
typedef struct triple_t {
    LSUP_Term *s;                   ///< Subject.
    LSUP_Term *p;                   ///< Predicate.
    LSUP_Term *o;                   ///< Object.
} LSUP_Triple;


/// Link type.
typedef enum {
    LSUP_LINK_INBOUND,              ///< Inbound link (sp).
    LSUP_LINK_OUTBOUND,             ///< Outbound link (po).
    LSUP_LINK_EDGE,                 ///< Edge link (so).
} LSUP_LinkType;


/** @brief The immediate neighborhood of terms connected to a term.
 *
 * This is a hash map whose each term is related to a set of one or more other
 * terms. The hash map is inside an opaque handle and is manipulated via the
 * `LSUP_link_map_*` functions.
 *
 * If the type of the link map is `LSUP_LINK_INBOUND`, the map keys
 * represent predicates and the sets related to them are the objects, and the
 * term associated to the link map is the object; if
 * `LSUP_LINK_OUTBOUND`, the keys represent predicates, the related sets
 * objects, and the associated term is the subject. If `LSUP_LINK_EDGE`, the
 * keys represent subjects and the related sets objects, and the associated
 * term is the predicate.
 */
typedef struct link_map LSUP_LinkMap;


/** @brief a set of unique terms.
 *
 * This is used to bulk-add terms to a link map.
 */
typedef struct hashmap LSUP_TermSet;


/*
 * External variables.
 */

/** @brief Compiled hash of default literal data type.
 */
extern uint32_t LSUP_default_dtype_key;

/** @brief URI validation pattern, compiled in #LSUP_init().
 */
extern regex_t *LSUP_uri_ptn;

/** @brief Default literal data type URI.
 *
 * Literal terms created with undefined data type will have it set to this
 * URI implicitly.
 */
extern LSUP_Term *LSUP_default_datatype;

/** @brief Global term cache.
 *
 * Stores frequently used terms, e.g. data type URIs.
 */
extern LSUP_TermSet *LSUP_term_cache;


/*
 * API functions.
 */

/** @brief Create a new term.
 *
 * This is a generic function; it is recommended to use specialized functions
 * such as #LSUP_term_new(), #LSUP_literal_new(), etc. as they have strict type
 * checks for the metadata parameter.
 *
 * @param type[in] Term type. One of #LSUP_TermType.
 *
 * @param data[in] Term data: textual URI, literal value without data type
 *  or langtag, etc. It may be NULL for IRI refs and BNodes, in which case a
 *  random identifier is generated.
 *
 * @param metadata[in] Namespace map (LSUP_NSMap *) for IRI refs; language tag
 * (LSUP_LangTag *) for language-tagged literals; or data type (LSUP_Term *)
 * for other literals. It may be NULL.
 *
 * @return New term, which must be freed with #LSUP_term_free after use; or
 *  NULL on error.
 */
LSUP_Term *
LSUP_term_new (LSUP_TermType type, const char *data, void *metadata);


/** @brief Placeholder term to use with LSUP_term_reset.
 */
#define TERM_DUMMY LSUP_term_new (LSUP_TERM_UNDEFINED, NULL, NULL)


/** @brief Shortcut to create an IRI reference.
 *
 * Must be freed with #LSUP_term_free.
 *
 * @param data[in] The URI string. If NULL, a UUID4-based URN is generated.
 *  This cannot be NULL if the nsm parameter is not NULL.
 *
 * @param nsm[in] Namespace map. If not NULL, a namespace-prefixed
 *  (#LSUP_TERM_NS_IRIREF) is created, otherwise a regular one
 *  (#LSUP_TERM_IRIREF).
 *
 * @return same as #LSUP_term_new().
 */
inline LSUP_Term *
LSUP_iriref_new (const char *data, LSUP_NSMap *nsm)
{
    return (
            nsm ? LSUP_term_new (LSUP_TERM_NS_IRIREF, data, nsm) :
            LSUP_term_new (LSUP_TERM_IRIREF, data, NULL));
}


/** @brief Create a new absolute IRI from a path relative to a root IRI.
 *
 * The term is always of type LSUP_TERM_IRIREF (i.e. not namespace-prefixed).
 *
 * If the provided IRI is already a fully qualified IRI (i.e. it has a prefix)
 * the result is semantically identical to the input.
 *
 * If the relative IRI begins with a '/', the resulting IRI is relative to the
 * web root of the root IRI. I.e. if a root IRI has a path after the webroot,
 * it is ignored.
 *
 * Otherwise, the resulting IRI is relative to the full root string.
 *
 * @param[in] root Root IRI that the new IRI should be relative to.
 *
 * @param[in] iri Term with an IRI relative to the webroot.
 *
 * @return New absolute IRI, or NULL if either term is not an IRI.
 */
LSUP_Term *
LSUP_iriref_absolute (const LSUP_Term *root, const LSUP_Term *iri);


/** @brief Create a new relative IRI from an absolute IRI and a web root IRI.
 *
 * This works with namespace-prefixed IRIs and returns a term of the same type
 * as the input.
 *
 * @param[in] iri Full IRI.
 *
 * @param[in] root Root IRI that the new IRI should be relative to.
 *
 * @return New IRI, or NULL if either term is not an IRI. If the input IRI is
 * not a path under the root IRI, the result will be identical to the input.
 */
LSUP_Term *
LSUP_iriref_relative (const LSUP_Term *root, const LSUP_Term *iri);


/** @brief Shortcut to create a literal term.
 *
 * Must be freed with #LSUP_term_free.
 *
 * @param data[in] The literal string.
 *
 * @param datatype[in] Data type URI string. If NULL, the default data type
 * (xsd:string) is used. The new term takes ownership of the pointer.
 *
 * @return same as #LSUP_term_new().
 */
inline LSUP_Term *
LSUP_literal_new (const char *data, LSUP_Term *datatype)
{ return LSUP_term_new (LSUP_TERM_LITERAL, data, datatype); }


/** @brief Shortcut to create a language-tagged literal term.
 *
 * Must be freed with #LSUP_term_free.
 *
 * @param data[in] The literal string.
 *
 * @param lang[in] Language tag string.
 *
 * @return same as #LSUP_term_new().
 */
inline LSUP_Term *
LSUP_lt_literal_new (const char *data, char *lang)
{ return LSUP_term_new (LSUP_TERM_LT_LITERAL, data, lang); }


/** @brief Shortcut to create a blank node.
 *
 * Must be freed with #LSUP_term_free.
 *
 * @param data[in] The BNode identifier.
 *
 * @return same as #LSUP_term_new().
 */
inline LSUP_Term *
LSUP_bnode_new (const char *data)
{ return LSUP_term_new (LSUP_TERM_BNODE, data, NULL); }


/** @brief Copy a term.
 *
 * @param[in] src The term to copy.
 *
 * @return A new duplicate term handle.
 */
LSUP_Term *
LSUP_term_copy (const LSUP_Term *src);


/** @brief Deserialize a buffer into a term.
 *
 * @param[in] sterm Buffer to convert into a term. It must be a valid
 *  serialized term from store or obtained with #LSUP_term_serialize().
 *
 * @return New term handle. It must be freed with #LSUP_term_free().
 */
LSUP_Term *
LSUP_term_new_from_buffer (const LSUP_Buffer *sterm);


/** @brief Serialize a term into a buffer.
 *
 * @param[in] sterm Term to convert into a buffer.
 *
 * @return New buffer handle. It must be freed with #LSUP_buffer_free().
 */
LSUP_Buffer *
LSUP_term_serialize (const LSUP_Term *term);


/** @brief Hash a buffer.
 */
LSUP_Key
LSUP_term_hash (const LSUP_Term *term);


/** @brief Compare two terms.
 *
 * The terms evaluate as equal if their hashes are equal—i.e. if they are
 * semantically equivalent.
 */
inline bool LSUP_term_equals (const LSUP_Term *term1, const LSUP_Term *term2)
{ return LSUP_term_hash (term1) == LSUP_term_hash (term2); }


void
LSUP_term_free (LSUP_Term *term);


/** @brief Namespace map of a IRI ref.
 *
 * @param[in] iri IRI reference handle.
 *
 * @return A pointer to the namespace map associated with the IRI. It is
 *  freed at program shutdown.
 */
LSUP_NSMap *
LSUP_iriref_nsm (const LSUP_Term *iri);


/** @brief Get the prefix portion of a IRI ref.
 *
 * @param[in] iri IRI reference handle.
 *
 * @return String containing the protocol and domain name part of the IRI. It
 *  should be freed after use.
 */
char *
LSUP_iriref_prefix (const LSUP_Term *iri);


/** @brief Get the path portion of a IRI ref.
 *
 * @param[in] iri IRI reference handle.
 *
 * @return String containing the path of the IRI relative to the web root. For
 * a URN, such as `urn:myns:myid`, it would be `myns:myid`. This string should
 * be freed after use.
 */
char *
LSUP_iriref_path (const LSUP_Term *iri);


/** @brief Get the fragment portion of a IRI ref.
 *
 * @param[in] iri IRI reference handle.
 *
 * @return String containing the fragment part of the IRI, or NULL if the IRI
 * contains no fragment. It should be freed after use.
 */
char *
LSUP_iriref_frag (const LSUP_Term *iri);

/*
 * TRIPLES
 */

/** @brief Create a new triple from three terms.
 *
 * Terms are NOT copied. To free them with the triple, use #LSUP_triple_free().
 * To only free the triple, use free().
 *
 * TODO Term types are not validated at the moment.
 *
 * @param[in] s Triple subject. It must be an IRIRef or BNode.
 *
 * @param[in] p Triple predicate. It must be an IRIRef.
 *
 * @param[in] o Triple object.
 *
 */
LSUP_Triple *
LSUP_triple_new(LSUP_Term *s, LSUP_Term *p, LSUP_Term *o);


/** @brief Dummy triple with NULL slots. It is not a valid triple.
 */
#define TRP_DUMMY LSUP_triple_new (NULL, NULL, NULL)


LSUP_Triple *
LSUP_triple_new_from_btriple (const LSUP_BufferTriple *sspo);


LSUP_BufferTriple *
LSUP_triple_serialize (const LSUP_Triple *spo);


/** @brief Initialize internal term pointers in a heap-allocated triple.
 *
 * Terms are NOT copied. To free them with the triple, use #LSUP_triple_free().
 * To only free the triple, use free().
 *
 * @param spo[in] Triple pointer to initialize.
 */
LSUP_rc
LSUP_triple_init (LSUP_Triple *spo, LSUP_Term *s, LSUP_Term *p, LSUP_Term *o);


/** @brief Free the internal pointers of a triple.
 *
 * @param spo[in] Triple to be freed.
 */
void
LSUP_triple_done (LSUP_Triple *spo);


/** @brief Free a triple and all its internal pointers.
 *
 * NOTE: If the term pointers are not to be freed (e.g. they are owned by a
 * back end), use a simple free(spo) instead of this.
 *
 * @param spo[in] Triple to be freed.
 */
void
LSUP_triple_free (LSUP_Triple *spo);


/** @brief Get triple by term position.
 *
 * Useful for looping over all terms.
 *
 * @param trp[in] Triple pointer.
 *
 * @param n[in] A number between 0÷2.
 *
 * @return Corresponding triple term or NULL if n is out of range.
 */
inline LSUP_Term *
LSUP_triple_pos (const LSUP_Triple *trp, LSUP_TriplePos n)
{
    if (n == TRP_POS_S) return trp->s;
    if (n == TRP_POS_P) return trp->p;
    if (n == TRP_POS_O) return trp->o;
    return NULL;
}


/** @brief Hash a triple.
 *
 * TODO This doesn't handle blank nodes correctly.
 */
inline LSUP_Key
LSUP_triple_hash (const LSUP_Triple *trp)
{
    LSUP_BufferTriple *strp = LSUP_triple_serialize (trp);
    LSUP_Key hash = LSUP_btriple_hash (strp);
    LSUP_btriple_free (strp);

    return hash;
}


/** @brief Create a new term set.
 *
 * @return New empty term set.
 */
LSUP_TermSet *
LSUP_term_set_new (void);


/** @brief Free a term set.
 *
 * @param[in] ts Term set handle.
 */
void
LSUP_term_set_free (LSUP_TermSet *ts);


/** @brief Add term to a term set.
 *
 * If the same term is already in the set, it is not replaced, and the existing
 * term's handle is made available in the `existing` variable. In this case,
 * the caller may want to free the passed term which has not been added.
 *
 * @param[in] tl Term set to be added to.
 *
 * @param[in] term Term to be added to the list. The term set will take
 * ownership of the term and free it when it's freed with
 * #LSUP_term_set_free()—only if the return code is LSUP_OK.
 *
 * @param[out] existing If not NULL, and if the term being added is a
 * duplicate, this variable will be populated with the existing term handle.
 *
 * @return LSUP_OK on success; LSUP_NOACTION if the term is duplicate;
 *  LSUP_MEM_ERR on memory error. Note: if not LSUP_OK, the caller is in charge
 *  of freeing the `term` handle.
 */
LSUP_rc
LSUP_term_set_add (LSUP_TermSet *ts, LSUP_Term *term, LSUP_Term **existing);


/** @brief Get a term from a term set.
 *
 * @param[in] ts Term set handle.
 *
 * @param[in] key Key for the queried term.
 *
 * @return The retrieved term if found, or NULL. The term must not be
 *  modified or freed.
 */
const LSUP_Term *
LSUP_term_set_get (LSUP_TermSet *ts, LSUP_Key key);


/** @brief Iterate trough a term set.
 *
 * @param[in] ts Term set handle.
 *
 * @param[in,out] i Iterator to be initially set to 0.
 *
 * @param[out] term Pointer to be populated with the next term on success. It
 *  may be NULL.
 *
 * @return LSUP_OK if the next term was retrieved; LSUP_END if the end of the
 *  set has been reached.
 */
LSUP_rc
LSUP_term_set_next (LSUP_TermSet *ts, size_t *i, LSUP_Term **term);


/** @brief New link map.
 *
 * The initial state of the returned list is: `{t: [NULL], tl: [NULL]}`
 *
 * Predicates and term lists can be added with #LSUP_link_map_add, and terms
 * can be added to a term list with #LSUP_term_list_add.
 *
 * @return a new empty predicate-object list.
 */
LSUP_LinkMap *
LSUP_link_map_new (LSUP_LinkType type);


/** @brief Free a link map.
 *
 * All arrays and term handles are recursively freed.
 *
 * @param[in] pol link map handle obtained with #LSUP_link_map_new().
 */
void
LSUP_link_map_free (LSUP_LinkMap *pol);


/// Return the link map type.
LSUP_LinkType
LSUP_link_map_type (const LSUP_LinkMap *map);


/** @brief Add a term - term set pair to a link map.
 *
 * If there is already a term set for the given term, items from the added term
 * are added to the existing term set (if not duplicated). Otherwise, the term
 * set handle is linked to the new term.
 *
 * In any case, the caller should not directly use the term and term set after
 * passing them to this function.
 *
 * @param[in] cm Link map handle obtained with #LSUP_link_map_new().
 *
 *  @param[in] t Term to be associated with the given object list. The
 *   link map structure takes ownership of the term.
 *
 * @param[in] ts term set to be associated with the given term. The link
 *  list structire takes ownership of the term set and the terms in it.
 *
 * @return LSUP_OK on success; LSUP_MEM_ERR on allocation error.
 */
LSUP_rc
LSUP_link_map_add (
        LSUP_LinkMap *cmap, LSUP_Term *term, LSUP_TermSet *tset);


/** @brief Create a new iterator to loop through a link map.
 *
 * @param[in] lmap Map handle to iterate.
 *
 * @param[in] ext External term to look for connections.
 */
LSUP_LinkMapIterator *
LSUP_link_map_iter_new (const LSUP_LinkMap *lmap, LSUP_Term *ext);


/// Free a link map iterator.
void
LSUP_link_map_iter_free (LSUP_LinkMapIterator *it);


/** @brief Iterate through a link map.
 *
 * Each call to this function yields a linked term and the related term set.
 *
 * @param[in] it Link map iterator obtained with #LSUP_link_map_iter_new().
 *
 * @param[out] lt Linked term returned.
 *
 * @param[out] ts Term set returned.
 *
 * @return LSUP_OK if a result was yielded; LSUP_END if the end of the link map
 *  has been reached.
 */
LSUP_rc
LSUP_link_map_next (
        LSUP_LinkMapIterator *it, LSUP_Term **lt, LSUP_TermSet **ts);


/**@brief Iterate over a link map and generate triples.
 *
 * Calling this function repeatedly builds triples for all the linked terms and
 * term sets in the map, based on a given related term.
 *
 * @param[in] it Link map iterator handle, obtained with
 *  #LSUP_link_map_iter_new().
 *
 * @param[in] term Term to relate to the link map.
 *
 * @param[in|out] spo Result triple. The triple handle must be pre-allocated
 *  (it may be TRP_DUMMY) and calls to this function will be set its memebers
 *  to term handles owned by the link map. If rc != LSUP_OK, the contents are
 *  undefined.
 *
 * @return LSUP_OK if a new triple was yielded; LSUP_END if the end of the loop
 *  has been reached; <0 on error.
 */
LSUP_rc
LSUP_link_map_triples (
        LSUP_LinkMapIterator *it, LSUP_Triple *spo);

#endif