volksdata/src/codec/codec_ttl.c

#include "volksdata/codec/codec_ttl.h"


/** @brief NT codec iterator.
 *
 * This iterator yields one or more triples at a time, one group per subject,
 * with the most compact form allowed by Turtle, e.g.
 *
 * ```
 * :s :p1 :o1, :o2, o3; :p2 :o4, :o5, <http://example.com/ext1> .
 * ```
 */
typedef struct {
    const VOLK_Codec *  codec;      ///< Codec that generated this iterator.
    const VOLK_Graph *  gr;         ///< Graph being encoded.
    VOLK_TermSet *      subjects;   ///< All subjects in the graph.
    size_t              s_cur;      ///< Term set cursor.
    VOLK_rc             rc;         ///< Internal return code.
    char *              s_str;      ///< Serialized subject block (output).
    char *              p_str;      ///< Serialized predicate block.
    char *              o_str;      ///< Serialized object block.
} VOLK_TTLCodecIterator;


/* * * Codec functions. * * */

static VOLK_rc
term_to_ttl (const VOLK_Term *term, char **out_p)
{
    VOLK_rc rc;
    char
        *tmp = NULL,
        *out = NULL,
        *metadata = NULL;
    size_t buf_len;

    VOLK_rc md_rc = VOLK_NORESULT;
    switch (term->type) {
        case VOLK_TERM_IRIREF:
            if (strcmp (term->data, VOLK_RDF_TYPE) == 0) {
                // Shorten RDF type
                buf_len = 2;
                out = realloc (*out_p, 2);
                if (UNLIKELY (!out)) return VOLK_MEM_ERR;
                out[0] = 'a';
                out[1] = '\0';
            } else {
                md_rc = VOLK_nsmap_denormalize_uri (term->data, &tmp);
                PRCCK (md_rc);
                if (md_rc == VOLK_NORESULT) {
                    // If URI counld not be shortened, add `<>`
                    // and copy term from the original.
                    out = realloc (*out_p, strlen (term->data) + 3);
                    if (UNLIKELY (!out)) return VOLK_MEM_ERR;
                    sprintf (out, "<%s>", term->data);
                } else {
                    // If URI was shortened, write it out without `<>` and
                    // use previously allocated data from denormalization.
                    // Free previous output pointer
                    free (*out_p);
                    out = tmp;
                }
            }
            rc = VOLK_OK;
            break;

        case VOLK_TERM_LITERAL:
            // Calculate string length.
            if (escape_lit (term->data, &tmp) != VOLK_OK)
                return VOLK_ERROR;
            buf_len = strlen (tmp) + 3; // Room for "" and terminator

            // Data type.
            bool shorten = false;
            if (
                term->datatype != 0
                && term->datatype != VOLK_default_datatype
            ) {
                md_rc = VOLK_nsmap_denormalize_uri (
                        term->datatype->data, &metadata);
                PRCCK (md_rc);
                unsigned padding = 0;

                // Shorten numeric and boolean types.
                if (strcmp (metadata, "xsd:integer") == 0) {
                    // TODO check for valid format.
                    shorten = true;
                } else if (strcmp (metadata, "xsd:double") == 0) {
                    // TODO check for valid format.
                    shorten = true;
                } else if (strcmp (metadata, "xsd:decimal") == 0) {
                    // TODO check for valid format.
                    shorten = true;
                } else if (strcmp (metadata, "xsd:boolean") == 0) {
                    // TODO check for valid format.
                    shorten = true;
                } else {
                    // Room for `^^<>` for FQURI, `^^` for NS URI
                    padding = md_rc == VOLK_NORESULT ? 4 : 2;
                }
                buf_len += strlen (metadata) + padding;
            }

            out = realloc (*out_p, buf_len);
            if (UNLIKELY (!out)) return VOLK_MEM_ERR;

            if (shorten) {
                strcpy (out, tmp);
            } else if (metadata) {
                char *fmt = (
                        md_rc == VOLK_NORESULT ? "\"%s\"^^<%s>"
                        : "\"%s\"^^%s");
                sprintf (out, fmt, tmp, metadata);
            }
            else {
                sprintf (out, "\"%s\"", tmp);
            }
            free (tmp);

            rc = VOLK_OK;

            break;

        case VOLK_TERM_LT_LITERAL:
            // Calculate string length.
            if (escape_lit (term->data, &tmp) != VOLK_OK)
                return VOLK_ERROR;
            buf_len = strlen (tmp) + 3; // Room for "" and terminator

            if (term->lang[0] != '\0') {
                metadata = strndup (term->lang, sizeof (VOLK_LangTag));
                buf_len += strlen (metadata) + 1; // Room for @
            }

            out = realloc (*out_p, buf_len);
            if (UNLIKELY (!out)) return VOLK_MEM_ERR;

            sprintf (out, "\"%s\"", tmp);
            free (tmp);

            // Add lang.
            if (metadata) out = strcat (strcat (out, "@"), metadata);

            rc = VOLK_OK;

            break;

        case VOLK_TERM_BNODE:
            out = realloc (*out_p, strlen (term->data) + 3);
            if (UNLIKELY (!out)) return VOLK_MEM_ERR;

            sprintf (out, "_:%s", term->data);
            rc = VOLK_OK;

            break;

        default:
            out = *out_p;  // This is considered garbage.
            log_error ("Invalid term type: %d", term->type);
            rc = VOLK_PARSE_ERR;
    }
    free (metadata);

    *out_p = out;
    return rc;
}


static void *
gr_to_ttl_init (const VOLK_Graph *gr)
{
    VOLK_TTLCodecIterator *it;
    CALLOC_GUARD (it, NULL);

    it->codec = &ttl_codec;
    it->gr = gr;
    it->subjects = VOLK_graph_unique_terms (gr, TRP_POS_S);
    // Sets the condition to build the prolog on 1st iteration.
    it->rc = VOLK_NORESULT;

    return it;
}


/// Build header and prolog.
static VOLK_rc
build_prolog (VOLK_TTLCodecIterator *it, char **res_p)
{
    char *res = fmt_header ("# ");

    const char ***nsm = VOLK_nsmap_dump ();
    char *ns_tpl = "@prefix %s: <%s> .\n";

    // Prefix map.
    for (size_t i = 0; nsm[i]; i++) {
        const char **ns = nsm[i];
        size_t old_len = strlen (res);
        size_t ns_len = strlen (ns[0]) + strlen (ns[1]) + strlen (ns_tpl);
        char *tmp = realloc (res, old_len + ns_len + 1);
        if (UNLIKELY (!tmp)) return VOLK_MEM_ERR;
        res = tmp;

        sprintf (res + old_len, ns_tpl, ns[0], ns[1]);
        free (ns);
    }
    free (nsm);

    // Base.
    char *base_uri_str = VOLK_graph_uri (it->gr)->data;
    char *base_stmt_tpl = "\n@base <%s> .\n\n";
    char *base_stmt = malloc (
            strlen (base_stmt_tpl) + strlen (base_uri_str) + 1);
    if (!UNLIKELY (base_stmt)) return VOLK_MEM_ERR;
    sprintf (base_stmt, base_stmt_tpl, base_uri_str);
    res = realloc (res, strlen (res) + strlen (base_stmt) + 1);
    if (!UNLIKELY (res)) return VOLK_MEM_ERR;
    res = strcat (res, base_stmt);
    free (base_stmt);

    *res_p = res;
    it->rc = VOLK_OK;

    return VOLK_OK;
}


/// Encode all the triples for a single subject.
static VOLK_rc
gr_to_ttl_iter (void *h, char **res_p) {
    VOLK_TTLCodecIterator *it = h;

    if (it->rc == VOLK_NORESULT) return build_prolog (it, res_p);

    VOLK_Term *s = NULL;
    char *res = *res_p;  // Result string will be reallocated.
    VOLK_rc rc = VOLK_term_set_next (it->subjects, &it->s_cur, &s);
    if (rc == VOLK_END) return rc;  // Return without logging error.
    RCCK (rc);  // Log error or warning for anything else.

    term_to_ttl (s, &res);

    VOLK_LinkMap *lmap = VOLK_graph_connections (
            it->gr, s, VOLK_LINK_OUTBOUND);

    VOLK_LinkMapIterator *lmit = VOLK_link_map_iter_new (lmap);
    VOLK_Term *p = NULL;
    VOLK_TermSet *o_ts = NULL;
    char *p_join = "\n    ";  // Newline & indent after subject.
    // Begin predicate loop.
    while (VOLK_link_map_next (lmit, &p, &o_ts) != VOLK_END) {
        // Add predicate representation.
        RCCK (term_to_ttl (p, &it->p_str));
        char *tmp = realloc (
                res, strlen (res) + strlen (it->p_str) + strlen (p_join) + 1);
        if (UNLIKELY (!tmp)) goto memfail;
        res = strcat (strcat (tmp, p_join), it->p_str);

        p_join = " ;\n    ";

        // Add objects for predicate.
        size_t i = 0;
        VOLK_Term *o = NULL;
        char *o_join = " ";
        while (VOLK_term_set_next (o_ts, &i, &o) != VOLK_END) {
            it->rc = term_to_ttl (o, &it->o_str);
            RCCK (it->rc);
            char *tmp = realloc (
                    res, strlen (res) + strlen (it->o_str) + strlen (o_join) + 1);
            if (UNLIKELY (!tmp)) goto memfail;
            res = strcat (strcat (tmp, o_join), it->o_str);
            o_join = " ,\n        ";  // Double indent for objects.
        }
    }

    char *s_sep = "\n.\n\n";  // Period goes on its own line for visibility.
    char *tmp = realloc (res, strlen (res) + strlen (s_sep) + 1);
    if (UNLIKELY (!tmp)) goto memfail;

    *res_p = strcat (tmp, s_sep);

    VOLK_link_map_iter_free (lmit);
    VOLK_link_map_free (lmap);

    return it->rc;

memfail:
    free (res);
    *res_p = NULL;
    return VOLK_MEM_ERR;
}


static void
gr_to_ttl_done (void *h)
{
    VOLK_TTLCodecIterator *it = h;
    VOLK_term_set_free (it->subjects);
    free (it->s_str);
    free (it->p_str);
    free (it->o_str);
    free (it);
}


const VOLK_Codec ttl_codec = {
    .name               = "Turtle",
    .mimetype           = "text/turtle",
    .extension          = "ttl",

    .encode_term        = term_to_ttl,

    .encode_graph_init  = gr_to_ttl_init,
    .encode_graph_iter  = gr_to_ttl_iter,
    .encode_graph_done  = gr_to_ttl_done,

    //.decode_term        = VOLK_ttl_parse_term,
    .decode_graph       = VOLK_ttl_parse_doc,
};