term.c

Go to the documentation of this file.

#include "volksdata/term.h"
 
#define MAX_VALID_TERM_TYPE     VOLK_TERM_BNODE /* For type validation. */
 
 
/*
 * Data structures.
 */

typedef struct keyed_term {
    VOLK_Key            key;        
    VOLK_Term *         term;       
} KeyedTerm;
 

typedef struct link {
    KeyedTerm *         term;       
    VOLK_TermSet *      tset;       
} Link;
 

struct link_map_iter {
    const VOLK_LinkMap *map;        
    size_t              i;          
    size_t              j;          
    const Link *        link;       
};
 
 
/*
 * A link map is thus nested:
 *
 * - A link map contains a hash map of Link instances (link).
 * - It also contains the single term that the other terms are related to
 *    (linked_t).
 * - Each Link contains a KeyedTerm (term) and a TermSet (tset).
 * - Each term set is a hash map of KeyedTerm instances.
 * - Each KeyedTerm contains a Term and its hash.
 */
typedef struct link_map {
    VOLK_LinkType       type;       
    VOLK_Term          *linked_t;   
    struct hashmap     *links;      
} VOLK_LinkMap;
 
 
/*
 * External variables.
 */
 
uint32_t VOLK_default_dtype_key = 0;
VOLK_Term *VOLK_default_ctx = NULL;
VOLK_Term *VOLK_default_datatype = NULL;
VOLK_TermSet *VOLK_term_cache = NULL;
 
 
/*
 * Static variables.
 */
 
// Characters not allowed in a URI string.
static const char *invalid_uri_chars = "<>\" {}|\\^`";

static const VOLK_TermType MIN_VALID_TYPE = VOLK_TERM_IRIREF;
static const VOLK_TermType MAX_VALID_TYPE = VOLK_TERM_BNODE;
 
/*
 * Static prototypes.
 */
 
static VOLK_rc
term_init (
        VOLK_Term *term, VOLK_TermType type, const char *data, void *metadata);
 
 
/*
 * Term set callbacks.
 */
 
static uint64_t
tset_hash_fn (
        const void *item, uint64_t _unused, uint64_t _unused2)
{
    (void) _unused;
    (void) _unused2;
 
    return ((const KeyedTerm *) item)->key;
}
 
 
static int
tset_cmp_fn (const void *a, const void *b, void *_unused)
{
    (void) _unused;
 
    return 
            ((const KeyedTerm *) a)->key -
            ((const KeyedTerm *) b)->key;
}
 
 
static void
tset_free_fn (void *item)
{ VOLK_term_free (((KeyedTerm *) item)->term); }
 
 
/*
 * Link map callbacks.
 */
 
static uint64_t
link_map_hash_fn (
        const void *item, uint64_t _unused, uint64_t _unused2)
{
    (void) _unused;
    (void) _unused2;
 
    return ((const Link *)item)->term->key;
}
 
 
static int
link_map_cmp_fn (const void *a, const void *b, void *_unused)
{
    (void) _unused;
 
    return
            ((const Link *)a)->term->key -
            ((const Link *)b)->term->key;
}
 
 
static void
link_map_free_fn (void *item)
{
    Link *link = item;
    VOLK_term_free (link->term->term);
    free (link->term);
    VOLK_term_set_free (link->tset);
}
 
 
 /*
 * Term API.
 */
 
VOLK_Term *
VOLK_term_new (
        VOLK_TermType type, const char *data, void *metadata)
{
    VOLK_Term *term;
    CALLOC_GUARD (term, NULL);
 
    if (UNLIKELY (term_init (
                    term, type, data, metadata) != VOLK_OK)) {
        free (term);
        return NULL;
    }
 
    return term;
}
 
 
VOLK_Term *
VOLK_term_copy (const VOLK_Term *src)
{
    switch (src->type) {
        case VOLK_TERM_IRIREF:
            return VOLK_iriref_new (src->data);
        case VOLK_TERM_LITERAL:
            return VOLK_literal_new (src->data, src->datatype);
        case VOLK_TERM_LT_LITERAL:
            return VOLK_lt_literal_new (src->data, (char *) src->lang);
        case VOLK_TERM_BNODE:
            return VOLK_term_new (VOLK_TERM_BNODE, src->data, src->metadata);
        default:
            log_error ("Unsupported term type: %d", src->type);
            return NULL;
    }
}
 

VOLK_Term *
VOLK_term_new_from_buffer (const VOLK_Buffer *sterm)
{
    if (UNLIKELY (!sterm)) return NULL;
 
    VOLK_TermType type;
    char *data;
    void *metadata = NULL;
 
    // Copy term type.
    size_t cplen = sizeof (type);
    char *cpcur = (char *)sterm->addr;
    memcpy (&type, cpcur, cplen);
 
    // Copy term data.
    cpcur += cplen;
    cplen = strlen (cpcur) + 1;
    data = malloc (cplen);
    NLNL (data);
    memcpy (data, cpcur, cplen);
 
    // If applicable, create term metadata.
    cpcur += cplen;
    if (type == VOLK_TERM_LITERAL) {
        if (strlen(cpcur) > 0)
            NLNL (metadata = (void *) VOLK_iriref_new (cpcur));
    } else if (type == VOLK_TERM_LT_LITERAL) {
        cplen = sizeof (VOLK_LangTag);
        metadata = malloc (cplen);
        NLNL (metadata);
        memcpy (metadata, cpcur, cplen);
    }
 
    VOLK_Term *ret = VOLK_term_new (type, data, metadata);
 
    free (data);
    if (type == VOLK_TERM_LT_LITERAL) free (metadata);
 
    return ret;
}
 
 
VOLK_Term *
VOLK_iriref_new_abs (const VOLK_Term *root, const VOLK_Term *iri)
{
    if (iri->type != VOLK_TERM_IRIREF) {
        log_error ("Provided path is not an IRI.");
        return NULL;
    }
    if (root->type != VOLK_TERM_IRIREF) {
        log_error ("Provided root is not an IRI.");
        return NULL;
    }
 
    char
        *data,
        *pfx = VOLK_iriref_prefix (iri);
 
    if (strlen (pfx) > 0) data = iri->data;
 
    else if (iri->data[0] == '/') {
        free (pfx);
 
        pfx = VOLK_iriref_prefix (root);
        data = malloc (strlen (iri->data) + strlen (pfx) + 1);
        if (!data) return NULL;
 
        sprintf (data, "%s%s", pfx, iri->data);
 
    } else {
        data = malloc (strlen (iri->data) + strlen (root->data) + 1);
        if (!data) return NULL;
 
        sprintf (data, "%s%s", root->data, iri->data);
    }
    free (pfx);
 
    VOLK_Term *ret = VOLK_iriref_new (data);
    if (data != iri->data) free (data);
 
    return ret;
}
 
 
VOLK_Term *
VOLK_iriref_new_rel (const VOLK_Term *root, const VOLK_Term *iri)
{
    if (iri->type != VOLK_TERM_IRIREF) {
        log_error ("Provided path is not an IRI.");
        return NULL;
    }
    if (root->type != VOLK_TERM_IRIREF) {
        log_error ("Provided root is not an IRI.");
        return NULL;
    }
 
    size_t offset = (
            strstr (iri->data, root->data) == iri->data ?
            strlen (root->data) : 0);
 
    return VOLK_iriref_new (iri->data + offset);
}
 
 
VOLK_Buffer *
VOLK_term_serialize (const VOLK_Term *term)
{
    /* The serialized data are a byte string (unsigned char *) formatted in
     * the following way:
     *
     * - (unsigned char) term->type
     * - (char *) NUL-delimited term->data
     * - (char *) serialized metadata as byte string
     *
     * All fields are cast to uchar. The first field is fixed, the
     * second and third are NUL-delimited, hence all fields are easily
     * identifiable.
     *
     * Metadata are:
     *
     * - For VOLK_TERM_IRIREF, no data. IRI info is calculated on demand.
     * - For VOLK_TERM_LITERAL, a `char` (`\1` or `\2`) indicating if a
     *   language tag is present, followed by the fully-qualified data type URI
     *   or the language tag, as a `NUL`-delimited string. For a `xsd:string`
     *   literal with no language, it is a zero-length string.
     * - For VOLK_TERM_BNODE, no data. Skolemization ID is calculated on
     *   deserialization.
     *
     * In serializing a term, the fact that two terms of different types may
     * be semantically identical must be taken into account. Specifically, a
     * VOLK_TERM_LT_LITERAL with no language tag is identical to a
     * VOLK_TERM_LITERAL of xsd:string type, made up of the same string. Such
     * terms must have identical serializations.
     */
 
    if (UNLIKELY (!term)) return NULL;
 
    VOLK_Buffer *sterm;
    CALLOC_GUARD (sterm, NULL);
 
    sterm->size = sizeof(VOLK_TermType) + strlen(term->data) + 1;
    NLNL (sterm->addr = malloc (sterm->size));
 
    // Copy term type.
    size_t offset = 0;
    size_t cplen = sizeof(term->type);
    memcpy (sterm->addr, &term->type, cplen);
 
    // Copy term data.
    offset += cplen;
    cplen = strlen (term->data) + 1;
    memcpy (sterm->addr + offset, term->data, cplen);
 
    // If applicable, copy literal metadata.
    offset += cplen;
    // Copy data type URI string or lang tag.
    if (term->type == VOLK_TERM_LITERAL) {
        // Non-language-tagged term.
        // Don't store default datatype (xsd:string).
        if (term->datatype == VOLK_default_datatype) {
            NLNL (sterm->addr = realloc (sterm->addr, ++sterm->size));
            sterm->addr[offset] = '\0';
        }
        else {
            cplen = strlen (term->datatype->data) + 1;
            sterm->size += cplen;
            NLNL (sterm->addr = realloc (sterm->addr, sterm->size));
            memcpy (
                sterm->addr + offset, term->datatype->data, cplen);
        }
    } else if (term->type == VOLK_TERM_LT_LITERAL) {
        // Language-tagged term.
        cplen = sizeof (VOLK_LangTag);
        sterm->size += cplen;
        NLNL (sterm->addr = realloc (sterm->addr, sterm->size));
        memcpy (sterm->addr + offset, &term->lang, sizeof (VOLK_LangTag));
    }
 
    return sterm;
}
 
 
VOLK_Key
VOLK_term_hash (const VOLK_Term *term)
{
    VOLK_Buffer *buf;
 
    if (UNLIKELY (!term)) buf = BUF_DUMMY;
    else buf = VOLK_term_serialize (term);
 
    VOLK_Key key = VOLK_buffer_hash (buf);
 
    VOLK_buffer_free (buf);
 
    return key;
}
 
 
void
VOLK_term_free (VOLK_Term *term)
{
    if (UNLIKELY (!term)) return;
 
    free (term->data);
    free (term);
}
 
 
char *
VOLK_iriref_prefix (const VOLK_Term *iri)
{
    if (iri->type != VOLK_TERM_IRIREF) {
        log_error ("Term is not a IRI ref type.");
        return NULL;
    }
 
    // if (iri->iri_info->prefix.size == 0) return NULL;
    VOLK_IRIInfo iri_info;
    RCNL (VOLK_parse_iri (iri->data, &iri_info));
 
    return strndup (
            iri->data + iri_info.prefix.offset, iri_info.prefix.size);
}
 
 
char *
VOLK_iriref_path (const VOLK_Term *iri)
{
    if (iri->type != VOLK_TERM_IRIREF) {
        log_error ("Term is not a IRI ref type.");
        return NULL;
    }
 
    // if (iri->iri_info->path.size == 0) return NULL;
    VOLK_IRIInfo iri_info;
    RCNL (VOLK_parse_iri (iri->data, &iri_info));
 
    return strndup (iri->data + iri_info.path.offset, iri_info.path.size);
}
 
 
char *
VOLK_iriref_frag (const VOLK_Term *iri)
{
    if (iri->type != VOLK_TERM_IRIREF) {
        log_error ("Term is not a IRI ref type.");
        return NULL;
    }
 
    // if (iri->iri_info->frag.size == 0) return NULL;
    VOLK_IRIInfo iri_info;
    RCNL (VOLK_parse_iri (iri->data, &iri_info));
 
    return strndup (iri->data + iri_info.frag.offset, iri_info.frag.size);
}
 
 
/*
 * Triple API.
 */
 
VOLK_Triple *
VOLK_triple_new(VOLK_Term *s, VOLK_Term *p, VOLK_Term *o)
{
    VOLK_Triple *spo = malloc (sizeof (*spo));
    if (!spo) return NULL;
 
    if (UNLIKELY (VOLK_triple_init (spo, s, p, o))) {
        free (spo);
        return NULL;
    }
 
    return spo;
}
 
 
VOLK_Triple *
VOLK_triple_new_from_btriple (const VOLK_BufferTriple *sspo)
{
    VOLK_Triple *spo = malloc (sizeof (*spo));
    if (!spo) return NULL;
 
    spo->s = VOLK_term_new_from_buffer (sspo->s);
    spo->p = VOLK_term_new_from_buffer (sspo->p);
    spo->o = VOLK_term_new_from_buffer (sspo->o);
 
    return spo;
}
 
 
VOLK_BufferTriple *
VOLK_triple_serialize (const VOLK_Triple *spo)
{
    VOLK_BufferTriple *sspo = malloc (sizeof (*sspo));
    if (!sspo) return NULL;
 
    sspo->s = VOLK_term_serialize (spo->s);
    sspo->p = VOLK_term_serialize (spo->p);
    sspo->o = VOLK_term_serialize (spo->o);
 
    return sspo;
}
 
 
VOLK_rc
VOLK_triple_init (VOLK_Triple *spo, VOLK_Term *s, VOLK_Term *p, VOLK_Term *o)
{
    /* FIXME TRP_DUMMY is a problem here.
    if (! VOLK_IS_IRI (s) && s->type != VOLK_TERM_BNODE) {
        log_error ("Subject is not of a valid term type: %d", s->type);
        return VOLK_VALUE_ERR;
    }
    if (! VOLK_IS_IRI (p)) {
        log_error ("Predicate is not of a valid term type: %d", p->type);
        return VOLK_VALUE_ERR;
    }
    */
 
    spo->s = s;
    spo->p = p;
    spo->o = o;
 
    return VOLK_OK;
}
 
 
void
VOLK_triple_done (VOLK_Triple *spo)
{
    if (UNLIKELY (!spo)) return;
 
    VOLK_term_free (spo->s);
    VOLK_term_free (spo->p);
    VOLK_term_free (spo->o);
}
 
 
void
VOLK_triple_free (VOLK_Triple *spo)
{
    if (UNLIKELY (!spo)) return;
 
    VOLK_term_free (spo->s);
    VOLK_term_free (spo->p);
    VOLK_term_free (spo->o);
 
    free (spo);
}
 
 
/*
 * Multi-add functions.
 */
 
VOLK_TermSet *
VOLK_term_set_new ()
{
    // Capacity of 4 is an arbitrary guess.
    VOLK_TermSet *ts = hashmap_new (
            sizeof (KeyedTerm), 4, VOLK_HASH_SEED, 0,
            tset_hash_fn, tset_cmp_fn, tset_free_fn, NULL);
    if (UNLIKELY (hashmap_oom (ts))) return NULL;
 
    return ts;
}
 
 
VOLK_rc
VOLK_term_set_add (
        VOLK_TermSet *ts, VOLK_Term *term, VOLK_Term **ins)
{
    VOLK_Hash key = VOLK_term_hash (term);
    KeyedTerm entry_s = {.key = key, .term = term};
 
    const KeyedTerm *ex = hashmap_get (ts, &entry_s);
    if (ex) {
        if (ins) *ins = ex->term;
        if (term != ex->term) VOLK_term_free (term);
        return VOLK_NOACTION;
    }
 
    hashmap_set (ts, &entry_s);
    if (hashmap_oom (ts)) {
        VOLK_term_free (term);
        return VOLK_MEM_ERR;
    }
 
    if (ins) *ins = term;
    return VOLK_OK;
}
 
 
const VOLK_Term *
VOLK_term_set_get (VOLK_TermSet *ts, VOLK_Key key)
{
    const KeyedTerm *entry = hashmap_get (ts, &(KeyedTerm){.key=key});
    if (entry) log_trace ("ID found for key %lx: %s", key, entry->term->data);
    else log_trace ("No ID found for key %lx.", key);
 
    return (entry) ? entry->term : NULL;
}
 
 
VOLK_rc
VOLK_term_set_next (VOLK_TermSet *ts, size_t *i, VOLK_Term **term)
{
    KeyedTerm *kt = NULL;
    if (!hashmap_iter (ts, i, (void **)&kt)) return VOLK_END;
 
    if (term) *term = kt->term;
 
    return VOLK_OK;
}
 
 
void
VOLK_term_set_free (VOLK_TermSet *ts)
{
    if (UNLIKELY (!ts)) return;
    hashmap_free (ts);
}
 
 
VOLK_rc
VOLK_term_set_remove (VOLK_TermSet *ts, VOLK_Term *term)
{
    KeyedTerm kt_s = {.key = VOLK_term_hash (term)};
 
    const KeyedTerm *res = hashmap_delete (ts, &kt_s);
 
    if (res) {
        VOLK_term_free (res->term);
        return VOLK_OK;
    }
 
    return VOLK_NOACTION;
}
 
 
VOLK_Term *
VOLK_term_set_pop (VOLK_TermSet *ts)
{
    const KeyedTerm *kt = NULL;
    size_t i = 0;
    if (!hashmap_iter (ts, &i, (void **)&kt)) return NULL;
 
    kt = hashmap_delete (ts, kt);
 
    return kt->term;
}
 
 
size_t
VOLK_term_set_size (VOLK_TermSet *ts)
{ return hashmap_count (ts); }
 
 
VOLK_LinkMap *
VOLK_link_map_new (const VOLK_Term *linked_term, VOLK_LinkType type)
{
    VOLK_LinkMap *lm;
    MALLOC_GUARD (lm, NULL);
    lm->type = type;
    lm->links = hashmap_new (
            sizeof (Link), 0, VOLK_HASH_SEED, 0,
            link_map_hash_fn, link_map_cmp_fn, link_map_free_fn, NULL);
    if (!linked_term) {
        log_error ("term must not be NULL.");
        free (lm);
        return NULL;
    }
    lm->linked_t = VOLK_term_copy (linked_term);
 
    return lm;
}
 
 
void
VOLK_link_map_free (VOLK_LinkMap *lm)
{
    hashmap_free (lm->links);
    VOLK_term_free (lm->linked_t);
    free (lm);
}
 
 
VOLK_LinkType
VOLK_link_map_type (const VOLK_LinkMap *map)
{ return map->type; }
 
 
// TODO Memory error handling.
VOLK_rc
VOLK_link_map_add (
        VOLK_LinkMap *lmap, VOLK_Term *term, VOLK_TermSet *tset)
{
    // Keyed term to look up the link term and insert it, if necessary.
    KeyedTerm entry_s = {.key=VOLK_term_hash (term), .term=term};
 
    const Link *ex = hashmap_get (lmap->links, &(Link){.term=&entry_s});
    if (ex) {
        // Add terms one by one to the existing term set.
        log_trace (
                "Linking term %s exists. Adding individual terms.",
                ex->term->term->data);
        size_t i = 0;
        KeyedTerm *kt;
        while (hashmap_iter (tset, &i, (void **)&kt)) {
            log_trace (
                    "Adding term %s to link %s",
                    kt->term->data, ex->term->term->data);
            if (hashmap_get (ex->tset, kt))
                // Term already exist, free the new one and move on.
                VOLK_term_free (kt->term);
            else
                // Insert KeyedTerm, the term set now owns the underlying term.
                hashmap_set (ex->tset, kt);
        }
        // Free link term that hasn't been used.
        VOLK_term_free (term);
    } else {
        // Add the new term and the termset wholesale.
        log_trace ("Adding new linking term %s.", term->data);
        // Allocate inserted member on heap, it will be owned by the map.
        KeyedTerm *ins;
        MALLOC_GUARD (ins, VOLK_MEM_ERR);
        memcpy (ins, &entry_s, sizeof (entry_s));
        Link link = {.term=ins, .tset=tset};
        hashmap_set (lmap->links, &link);
    }
 
    return VOLK_OK;
}
 
 
VOLK_LinkMapIterator *
VOLK_link_map_iter_new (const VOLK_LinkMap *lmap)
{
    VOLK_LinkMapIterator *it;
    CALLOC_GUARD (it, NULL);
    it->map = lmap;
 
    return it;
}
 
 
// This leaves the link and link map references intact.
void
VOLK_link_map_iter_free (VOLK_LinkMapIterator *it) { free (it); }
 
 
VOLK_rc
VOLK_link_map_next (
        VOLK_LinkMapIterator *it, VOLK_Term **lt, VOLK_TermSet **ts)
{
    if (!hashmap_iter (it->map->links, &it->i, (void **)&it->link))
        return VOLK_END;
 
    *lt = it->link->term->term;
    *ts = it->link->tset;
 
    return VOLK_OK;
}
 
 
// TODO dismantle if the only triple generator is for the graph.
VOLK_rc
VOLK_link_map_triples (
        VOLK_LinkMapIterator *it, VOLK_Triple *spo)
{
    // Assign external (related) term.
    if (it->map->type == VOLK_LINK_INBOUND)
        spo->o = it->map->linked_t;
    else if (it->map->type == VOLK_LINK_OUTBOUND)
        spo->s = it->map->linked_t;
    else spo->p = it->map->linked_t;
 
    KeyedTerm *kt;
 
    // If we are already handling a link, continue the internal loop.
    if (it->link) goto int_loop;
ext_loop:
    // Advance external counter and start new internal loop.
    it->j = 0;
    if (!hashmap_iter (it->map->links, &it->i, (void **)&it->link))
        return VOLK_END;
int_loop:
    // If end of the term set is reached, start with a new linking term.
    if (!hashmap_iter (it->link->tset, &it->j, (void **)&kt)) goto ext_loop;
 
    // Continue pulling from term set.
    // Assign linking term.
    if (it->map->type == VOLK_LINK_EDGE) spo->s = it->link->term->term;
    else spo->p = it->link->term->term;
 
    // Assign term in term set.
    if (it->map->type == VOLK_LINK_INBOUND) spo->s = kt->term;
    else spo->o = kt->term;
 
    return VOLK_OK;
}
 
 
/*
 * Static functions.
 */
 
static VOLK_rc
term_init (
        VOLK_Term *term, VOLK_TermType type,
        const char *data, void *metadata)
{
    // Exit early if environment is not initialized.
    // EXCEPT for IRIRef which is used inside of VOLK_init().
    if (!VOLK_env_is_init && type != VOLK_TERM_IRIREF)
        return VOLK_ENV_ERR;
 
    // Undefined type. Make quick work of it.
    if (type == VOLK_TERM_UNDEFINED) {
        term->type = type;
        if (data) {
            term->data = malloc (strlen (data) + 1);
            if (UNLIKELY (!term->data)) return VOLK_MEM_ERR;
            strcpy (term->data, data);
        }
        return VOLK_OK;
    }
 
    if (type < MIN_VALID_TYPE || type > MAX_VALID_TYPE) {
        log_error ("%d is not a valid term type.", type);
        return VOLK_VALUE_ERR;
    }
 
    term->type = type;
 
    if (data) {
        // Validate IRI.
        if (term->type == VOLK_TERM_IRIREF) {
            if (strpbrk (data, invalid_uri_chars) != NULL) {
                log_warn (
                        "Characters %s are not valid in a URI. Got: %s\n",
                        invalid_uri_chars, data);
#if 0
                // TODO This causes W3C TTL test #29 to fail. Remove?
                return VOLK_VALUE_ERR;
#endif
            }
        }
        term->data = strdup (data);
 
    } else {
        // No data. Make up a random UUID or URI if allowed.
        if (type == VOLK_TERM_IRIREF || type == VOLK_TERM_BNODE) {
            uuid_t uuid;
            uuid_generate_random (uuid);
 
            uuid_str_t uuid_str;
            uuid_unparse_lower (uuid, uuid_str);
 
            if (type == VOLK_TERM_IRIREF) {
                term->data = malloc (UUID4_URN_SIZE);
                snprintf (
                        term->data, UUID4_URN_SIZE, "urn:uuid:%s", uuid_str);
            } else term->data = strdup (uuid_str);
        } else {
            log_error ("No data provided for term.");
            return VOLK_VALUE_ERR;
        }
    }
 
    if (term->type == VOLK_TERM_LT_LITERAL) {
        if (!metadata) {
            log_warn ("Lang tag is NULL. Creating a non-tagged literal.");
            term->type = VOLK_TERM_LITERAL;
        } else {
            // FIXME metadata should be const all across.
            char *lang_str = (char *) metadata;
            log_trace ("Lang string: '%s'", lang_str);
            // Lang tags longer than 7 characters will be truncated.
            strncpy(term->lang, lang_str, sizeof (term->lang) - 1);
            if (strlen (term->lang) < 1) {
                log_error ("Lang tag cannot be an empty string.");
                return VOLK_VALUE_ERR;
            }
            term->lang[7] = '\0';
        }
    }
 
    // This also captures a LT_LITERAL with no lang tag.
    if (term->type == VOLK_TERM_LITERAL) {
        term->datatype = metadata ? metadata : VOLK_default_datatype;
 
        if (term->datatype->type != VOLK_TERM_IRIREF) {
            log_error (
                    "Literal data type is not an IRI: %s",
                    term->datatype->data);
 
            return VOLK_VALUE_ERR;
        }
 
        PRCCK (VOLK_term_set_add (
                    VOLK_term_cache, term->datatype, &term->datatype));
 
    } else if (term->type == VOLK_TERM_BNODE) {
        // TODO This is not usable for global skolemization.
        term->bnode_id = VOLK_HASH (
                term->data, strlen (term->data) + 1, VOLK_HASH_SEED);
    }
 
    return VOLK_OK;
}
 
 
VOLK_rc
VOLK_parse_iri (char *iri_str, VOLK_IRIInfo *iri_info) {
    char *cur = iri_str;
    size_t iri_len = strlen (iri_str);
    MatchCoord tmp = {};  // Temporary storage for capture groups
 
    memset (iri_info, 0, sizeof (*iri_info));
    //log_debug ("Parsing IRI: %s", iri_str);
    // #2: ([^:/?#]+)
    while (
            *cur != ':' && *cur != '/' && *cur != '?'
            && *cur != '#' && *cur != '\0') {
        tmp.size++;
        cur++;
    }
 
    // Non-capturing: (?([^:/?#]+):)?
    if (tmp.size > 0 && *cur == ':') {
        // Got capture groups #2 and #3. Store them.
        iri_info->scheme.offset = 0;
        iri_info->scheme.size = tmp.size;
        cur++;
        //log_debug ("Group #2: %lu, %lu", coord[2].offset, coord[2].size);
    } else cur = iri_str;  // Backtrack if no match.
 
    // Non-capturing: (?//([^/?#]*))?
    if (*cur == '/' && *(cur + 1) == '/') {
        cur += 2;
        tmp.offset = cur - iri_str;
        tmp.size = 0;
 
        // #3: ([^/?#]*)
        while (*cur != '/' && *cur != '?' && *cur != '#' && *cur != '\0') {
            tmp.size++;
            cur++;
        }
        iri_info->auth.offset = tmp.offset;
        iri_info->auth.size = tmp.size;
        //log_debug ("Group #3: %lu, %lu", coord[3].offset, coord[3].size);
    }
 
    // Capture group 1.
    iri_info->prefix.offset = 0;
    iri_info->prefix.size = cur - iri_str;
    //log_debug ("Group #1: %lu, %lu", coord[1].offset, coord[1].size);
 
    tmp.offset = cur - iri_str;
    tmp.size = 0;
 
    iri_info->path.offset = tmp.offset;
    iri_info->path.size = iri_len - tmp.offset;
    //log_debug ("Group #4: %lu, %lu", coord[4].offset, coord[4].size);
 
    // Non-capturing: (?[^?#]*)
    while (*cur != '?' && *cur != '#' && *cur != '\0') {
        tmp.size++;
        cur++;
    }
 
    // Non-capturing: (?\?([^#]*))
    if (*cur == '?') {
        // 5: ([^#]*)
        tmp.offset = ++cur - iri_str;
        tmp.size = 0;
        while (*cur != '#' && *cur != '\0') {
            tmp.size++;
            cur++;
        }
 
        if (tmp.size > 0) {
            // Got capture group #5.
            iri_info->query.offset = tmp.offset;
            iri_info->query.size = tmp.size;
            //log_debug ("Group #5: %lu, %lu", coord[5].offset, coord[5].size);
        }
    }
 
    // Non-capturing: (?#(.*))?
    if (*cur == '#') {
        // #6: (.*)
        iri_info->frag.offset = ++cur - iri_str;
        iri_info->frag.size = iri_str + iri_len - cur;
        //log_debug ("Group #6: %lu, %lu", coord[6].offset, coord[6].size);
    }
 
    /* TODO add error cases.
    if (UNLIKELY (rc != VOLK_OK)) {
        log_error ("Error matching URI pattern.");
 
        return VOLK_VALUE_ERR;
    }
    */
 
    return VOLK_OK;
}
 
 
/*
 * Extern inline functions.
 */
 
VOLK_Key VOLK_term_hash (const VOLK_Term *term);
VOLK_Term *VOLK_iriref_new (const char *data);
VOLK_Term *VOLK_iriref_new_ns (const char *data);
VOLK_Term *VOLK_literal_new (const char *data, VOLK_Term *datatype);
VOLK_Term *VOLK_lt_literal_new (const char *data, char *lang);
VOLK_Term *VOLK_bnode_new (const char *data);
bool VOLK_term_equals (const VOLK_Term *term1, const VOLK_Term *term2);
VOLK_Term *VOLK_triple_pos (const VOLK_Triple *trp, VOLK_TriplePos n);
VOLK_Key VOLK_triple_hash (const VOLK_Triple *trp);