Annotation-System/hmmlib2/hmmlib.h

#ifndef HMMLIB_H_
#define HMMLIB_H_
#include <stddef.h>
#include <stdio.h>

// Data structures

typedef struct {
    char* name;
    char* role;
} HMML_Credit;

typedef struct {
    char* key;
    char* value;
} HMML_VideoCustomMetaData;

typedef struct {
    char* stream_platform;
    char* project;
    char* title;
    char* vod_platform;
    char* id;
    char* output;
    char* template;
    char* medium;
    char* number;

    HMML_Credit* credits;
    size_t credit_count;

    HMML_Credit* uncredits;
    size_t uncredit_count;

    HMML_VideoCustomMetaData* custom;
    size_t custom_count;

} HMML_VideoMetaData;

typedef struct {
    char* site;
    char* page;
    char* url;
    char* title;
    char* article;
    char* author;
    char* editor;
    char* publisher;
    char* isbn;
    int offset;
} HMML_Reference;

typedef enum {
    HMML_CATEGORY,
    HMML_MEMBER,
    HMML_PROJECT,

    HMML_MARKER_COUNT,
} HMML_MarkerType;

typedef struct {
    HMML_MarkerType type;
    char* marker;
    char* parameter;
    char* episode;
    int offset;
} HMML_Marker;

typedef struct {
    _Bool present;
    int id;
    char* author;
} HMML_Quote;

typedef struct {
    int line;

    int h, m, s;

    char* text;
    char* author;

    HMML_Reference* references;
    size_t reference_count;

    HMML_Marker* markers;
    size_t marker_count;

    HMML_Quote quote;
} HMML_Timestamp;

typedef struct {
    int line;
    int col;
    char* message;
} HMML_Error;

typedef struct {
    _Bool              well_formed;
    HMML_VideoMetaData metadata;
    HMML_Timestamp*    timestamps;
    size_t             timestamp_count;
    HMML_Error         error;
    void*              free_list; // implementation detail
} HMML_Output;

// Functions

HMML_Output hmml_parse (const char* string);
void        hmml_free  (HMML_Output* output);

// Version

extern const struct HMML_Version {
    int Major, Minor, Patch;
} hmml_version;

#endif

#ifdef HMMLIB_IMPLEMENTATION

#include <setjmp.h>
#include <string.h>
#include <stdarg.h>
#include <stdlib.h>
#include <stdint.h>

#define HSTX(x) x, sizeof(x)-1
#define HSTR(x) (const struct _hmml_str){ HSTX(x) }

#ifndef MALLOC
    #define MALLOC malloc
#endif

#ifndef REALLOC
    #define REALLOC realloc
#endif

#ifndef countof
    #define countof(x) (sizeof(x)/sizeof(*x))
#endif

#define _hmml_debug(...)
//#define _hmml_debug printf

struct _hmml_parser {
    HMML_Output out;
    const char* mem;
    const char* cursor;
    jmp_buf     err_buf;
    uintptr_t*  free_list;
    int         line;
};

struct _hmml_str {
    const char*  ptr;
    size_t len;
};

// memory management boilerplate stuff

static void* _hmml_store_ptr(struct _hmml_parser* p, void* input)
{
    uintptr_t* ptr;
    if(p->free_list) {
        ptr = p->free_list;
        if(ptr[1] + 1 == ptr[0]) {
            size_t n = ptr[0] << 1;
            ptr = REALLOC(ptr, n * sizeof(uintptr_t));
            ptr[0] = n;
        }
        ptr[ptr[1]] = (uintptr_t)input;
        ptr[1]++;
    } else {
        ptr = MALLOC(8 * sizeof(uintptr_t));
        ptr[0] = 8;
        ptr[1] = 3;
        ptr[2] = (uintptr_t)input;
    }
    p->free_list = ptr;
    return input;
}

static char* _hmml_persist_str(struct _hmml_parser* p, const struct _hmml_str str)
{
    char* mem = MALLOC(str.len+1);
    memcpy(mem, str.ptr, str.len);
    mem[str.len] = '\0';
    return _hmml_store_ptr(p, mem);
}

static void _hmml_persist_array_fn(struct _hmml_parser* p, void** out, size_t* out_count, void* in, size_t in_size)
{
    void* base;
    if(!*out) {
        base = MALLOC(in_size + sizeof(size_t));
        _hmml_store_ptr(p, base);
        *(size_t*)base = p->free_list[1]-1;
    } else {
        base = (char*)(*out) - sizeof(size_t);
        base = REALLOC(base, (*out_count + 1) * in_size + sizeof(size_t));
        size_t free_list_off = *(size_t*)base;
        p->free_list[free_list_off] = (intptr_t)base;
    }

    *out = (char*)base + sizeof(size_t);
    memcpy((char*)*out + (*out_count * in_size), in, in_size);
    ++(*out_count);
}

#define _hmml_persist_array(p, out, out_count, in) \
    _hmml_persist_array_fn((p), (void**)(out), (out_count), &(in), sizeof(in))

// error handling

#define _hmml_err(p, fmt, ...) \
    _hmml_err_fn((p), fmt "\n", ##__VA_ARGS__)

__attribute__((noreturn))
static void _hmml_err_fn(struct _hmml_parser* p, const char* fmt, ...)
{
    static char error_buf[4096];

    va_list va;
    va_start(va, fmt);
    int n = vsnprintf(error_buf, sizeof(error_buf), fmt, va);
    va_end(va);

    int line = 1, col = 1;
    for(const char* ptr = p->mem; ptr != p->cursor; ++ptr) {
        if(*ptr == '\n') {
            ++line;
            col = 1;
        } else {
            ++col;
        }
    }

    p->out.error.message = _hmml_persist_str(p, (struct _hmml_str){ error_buf, n });
    p->out.error.line = line;
    p->out.error.col = col;

    longjmp(p->err_buf, 1);
}

// actual parsing stuff

static void _hmml_skip_ws(struct _hmml_parser* p)
{
    for(;;) {
        uint8_t c = *p->cursor;
        if(c && c <= ' ') {
            if(c == '\n') {
                ++p->line;
            }
            ++p->cursor;
        } else {
            break;
        }
    }
}

static _Bool _hmml_str_eq(struct _hmml_str a, struct _hmml_str b)
{
    return a.len == b.len && memcmp(a.ptr, b.ptr, a.len) == 0;
}

static _Bool _hmml_unesc(char in, char* out)
{
    if(strchr("[]:@~\\\"", in)) {
        *out = in;
        return 1;
    } else {
        return 0;
    }
}

static char* _hmml_read_attr(struct _hmml_parser* p, char* mem, size_t mem_size, _Bool break_on_punct)
{
    const char* src = p->cursor;
    char* dst = mem;

    if(*src == '"') {
        ++src;
        while(*src && *src != '"' && (size_t)(src - p->cursor) < mem_size) {
            char converted;
            if(*src == '\\' && _hmml_unesc(src[1], &converted)) {
                *dst++ = converted;
                src += 2;
            } else {
                *dst++ = *src++;
            }
        }

        if(*src != '"') {
            _hmml_err(p, "Partially quoted attribute");
        }

        *dst = '\0';
        p->cursor = src+1;
    } else {
        const char* breaks = break_on_punct
            ? " ]\r\n\t:,'-.#=[\\?!…()\"%"
            : " ]\r\n\t"
            ;

        size_t n = strcspn(src, breaks);
        if(n >= mem_size) {
            _hmml_err(p, "Attribute [%.10s...] too long", p->cursor);
        }
        memcpy(dst, src, n);
        dst += n;
        *dst = '\0';
        p->cursor += n;
    }

    return dst;
}

static void _hmml_read_kv(struct _hmml_parser* p, struct _hmml_str* key, struct _hmml_str* val)
{
    static char key_memory[64];
    static char val_memory[1024];

    size_t key_len = strcspn(p->cursor, " \r\n\t=");
    if(key_len >= sizeof(key_memory)) {
        _hmml_err(p, "Attribute key [%.10s...] too long", p->cursor);
    }

    memcpy(key_memory, p->cursor, key_len);
    key_memory[key_len] = '\0';
    p->cursor += key_len;

    _hmml_skip_ws(p);

    if(*p->cursor != '=') {
        _hmml_err(p, "Expected '=', got [%.3s]", p->cursor);
    }

    ++p->cursor;

    char* end = _hmml_read_attr(p, val_memory, sizeof(val_memory), 0);

    _hmml_debug("read kv [%s] = [%s]\n", key_memory, val_memory);

    key->ptr = key_memory;
    key->len = key_len;

    val->ptr = val_memory;
    val->len = end - val_memory;
}

static HMML_Marker _hmml_parse_marker(struct _hmml_parser* p)
{
    static char marker_mem[4096];

    // the extended markers are inside [ ] and can contain parameters
    _Bool extended = *p->cursor == '[';
    if(extended) {
        ++p->cursor;
    }

    HMML_Marker marker = {
        .offset = -1,
    };

    char c = *p->cursor;
    if(c == '~') {
        marker.type = HMML_PROJECT;
    } else if(c == '@') {
        marker.type = HMML_MEMBER;
    } else if(c == ':') {
        marker.type = HMML_CATEGORY;
    } else {
        _hmml_err(p, "Unknown marker type");
    }

    ++p->cursor;

    char* end = _hmml_read_attr(p, marker_mem, sizeof(marker_mem), !extended);
    marker.marker = _hmml_persist_str(p, (struct _hmml_str){ marker_mem, end - marker_mem });

    if(extended) {
        _hmml_skip_ws(p);

        if(*p->cursor == '#') {
            ++p->cursor;
            size_t n = strcspn(p->cursor, " ");
            marker.episode = _hmml_persist_str(p, (struct _hmml_str){ p->cursor, n });
            p->cursor += n + 1;
        }

        if(*p->cursor != ']') {
            const char* end = p->cursor;

            for(;;) {
                if(!*end) {
                    break;
                }

                char converted;
                if(*end == '\\' && _hmml_unesc(end[1], &converted)) {
                    end += 2;
                } else if(*end == ']'){
                    break;
                } else {
                    ++end;
                }
            }

            marker.parameter = _hmml_persist_str(p, (struct _hmml_str){ p->cursor, end - p->cursor });
            p->cursor = end;
        }

        if(*p->cursor != ']') {
            _hmml_err(p, "Expected ']'");
        }
        ++p->cursor;
    }

    return marker;
}

static HMML_Reference _hmml_parse_ref(struct _hmml_parser* p)
{
    HMML_Reference ref = {
        .offset = -1,
    };

    struct str_attr {
        struct _hmml_str str;
        char** dest;
    } str_attrs[] = {
        { HSTR("site")     , &ref.site },
        { HSTR("page")     , &ref.page },
        { HSTR("url")      , &ref.url },
        { HSTR("title")    , &ref.title },
        { HSTR("article")  , &ref.article },
        { HSTR("author")   , &ref.author },
        { HSTR("editor")   , &ref.editor },
        { HSTR("publisher"), &ref.publisher },
        { HSTR("isbn")     , &ref.isbn },
    };

    for(;;) {
next_attr:
        _hmml_skip_ws(p);

        if(*p->cursor == ']') {
            ++p->cursor;
            break;
        }

        struct _hmml_str key, value;
        _hmml_read_kv(p, &key, &value);

        for(size_t i = 0; i < countof(str_attrs); ++i) {
            struct str_attr* s = str_attrs + i;
            if(_hmml_str_eq(key, s->str)) {
                *s->dest = _hmml_persist_str(p, value);
                goto next_attr;
            }
        }

        _hmml_err(p, "Unknown reference attribute");
    }

    return ref;
}

static void _hmml_parse_timecode(struct _hmml_parser* p, HMML_Timestamp* ts)
{
    unsigned int h = 0, m = 0, s = 0;
    int offset = 0;
    int count = sscanf(p->cursor, "[%u:%u%n", &m, &s, &offset);

    if(count < 2) {
        _hmml_err(p, "Unable to parse timecode");
    }

    p->cursor += offset;
    char c = *p->cursor;

    if(c == ':') {
        unsigned int tmp;
        offset = 0;
        if(sscanf(p->cursor, ":%u]%n", &tmp, &offset) != 1 || offset == 0) {
            _hmml_err(p, "Unable to parse 3-part timecode");
        }

        h = m;
        m = s;
        s = tmp;

        p->cursor += offset;

    } else if(c != ']') {
        _hmml_err(p, "Unable to parse timecode");
    } else {
        ++p->cursor;
    }

    if(s >= 60) {
        _hmml_err(p, "Seconds cannot exceed 59");
    }

    if(m >= 60) {
        _hmml_err(p, "Minutes cannot exceed 59");
    }

    ts->h = h;
    ts->m = m;
    ts->s = s;
}

static void _hmml_store_marker(struct _hmml_parser* p, HMML_Timestamp* ts, char** out, char* text_mem, size_t text_mem_size)
{
    HMML_Marker m = _hmml_parse_marker(p);
    m.offset = (*out) - text_mem;
    _hmml_persist_array(p, &ts->markers, &ts->marker_count, m);

    const char* marker_text = m.parameter
        ? m.parameter
        : m.marker
        ;

    size_t text_len = strlen(marker_text);
    if((*out) + text_len > text_mem + text_mem_size) {\
        _hmml_err(p, "Not enough text memory");\
    }
    memcpy(*out, marker_text, text_len);
    *out += text_len;
}

static size_t _hmml_parse_text(struct _hmml_parser* p, HMML_Timestamp* ts)
{
    static char text_mem[4096];
    char* out = text_mem;

    memset(text_mem, 0, sizeof(text_mem));

    for(;;) {
        size_t n = strcspn(p->cursor, "\\\n\r[]:@~");
        char c = p->cursor[n];

        if(out + n > text_mem + sizeof(text_mem)) {\
            _hmml_err(p, "Not enough text memory");\
        }
        memcpy(out, p->cursor, n);

        p->cursor += n;
        out += n;

        if(c == '\0') {
            _hmml_err(p, "Unexpected EOF");
        }

        else if(c == ']') {
            ++p->cursor;
            break;
        }

        else if(c == '\\') {
            char converted;
            if(_hmml_unesc(p->cursor[1], &converted)) {
                *out++ = converted;
                p->cursor += 2;
            } else {
                *out++ = '\\';
                p->cursor++;
            }
        }

        else if(c == '\n' || c == '\r') {
            ++p->cursor;
        }

        else if(c == '[') {
            if(strncmp(p->cursor + 1, "ref", 3) == 0) {
                p->cursor += 4;
                HMML_Reference ref = _hmml_parse_ref(p);
                ref.offset = out - text_mem;
                _hmml_persist_array(p, &ts->references, &ts->reference_count, ref);
            } else {
                _hmml_store_marker(p, ts, &out, text_mem, sizeof(text_mem));
            }
        }

        // it is a @ ~ or : marker without parameters
        else {
            // if next char is a space, or prev char is not a space*, then it can't be a marker
            // * unless it's the first char
            if(strchr(" \t\r\n", p->cursor[1]) || !(out == text_mem || strchr(" \t\r\n", p->cursor[-1]))) {
                *out++ = c;
                ++p->cursor;
            } else {
                _hmml_store_marker(p, ts, &out, text_mem, sizeof(text_mem));
            }
        }

        if((size_t)(out - text_mem) >= sizeof(text_mem)) {
            _hmml_err(p, "Not enough text memory");
        }
    }

    // trim trailing whitespace
    while(out > text_mem && (uint8_t)(out[-1]) <= ' ') {
        out[-1] = '\0';
        --out;
    }

    size_t text_size = out - text_mem;
    ts->text = _hmml_persist_str(p, (struct _hmml_str){ text_mem, text_size });

    return text_size;
}

static void _hmml_parse_quote(struct _hmml_parser* p, HMML_Timestamp* ts)
{
    char member[256];
    int id;
    int off = 0;

    if(sscanf(p->cursor, "[quote %255s %d]%n", member, &id, &off) == 2 && off) {
        ts->quote.present = 1;
        ts->quote.id = id;
        ts->quote.author = _hmml_persist_str(p, (struct _hmml_str){ member, strlen(member) });
    } else if(sscanf(p->cursor, "[quote %d]%n", &id, &off) == 1 && off) {
        ts->quote.present = 1;
        ts->quote.id = id;
    } else {
        _hmml_err(p, "Unable to parse quote");
    }

    p->cursor += off;
}

static void _hmml_parse_timestamps(struct _hmml_parser* p)
{
    for(;;) {
        _hmml_skip_ws(p);

        if(*p->cursor == '\0') {
            _hmml_err(p, "Unexpected EOF");
        }

        if(strncmp(p->cursor, "[/video]", 8) == 0) {
            break;
        }

        HMML_Timestamp ts = {
            .line = p->line
        };

        _hmml_parse_timecode(p, &ts);

        if(*p->cursor != '[') {
            _hmml_err(p, "Expected '['");
        }

        if(p->cursor[1] == '@') {
            HMML_Marker m = _hmml_parse_marker(p);
            ts.author = m.marker;
        }

        ++p->cursor;

        int text_len = _hmml_parse_text(p, &ts);

        if(p->cursor[0] == '[' && p->cursor[1] == ':') {
            ++p->cursor;
            do {
                HMML_Marker m = _hmml_parse_marker(p);
                _hmml_persist_array(p, &ts.markers, &ts.marker_count, m);
                _hmml_skip_ws(p);
                if(*p->cursor != ':' && *p->cursor != ']') {
                    _hmml_err(p, "Unterminated post-text category node");
                }
            } while(*p->cursor == ':');
            ++p->cursor;
        }

        if(p->cursor[0] == '[' && p->cursor[1] == 'q') {
            _hmml_parse_quote(p, &ts);
        }

        // convert all markers to lowercase, fix any out of range offsets
        for(size_t i = 0; i < ts.marker_count; ++i) {
            HMML_Marker* m = ts.markers + i;
            for(char* c = m->marker; *c; ++c) {
                if(*c >= 'A' && *c <= 'Z') {
                    *c = (*c - ('A' - 'a'));
                }
            }
            if(m->offset > text_len) {
                m->offset = text_len;
            }
        }

        for(size_t i = 0; i < ts.reference_count; ++i) {
            HMML_Reference* ref = ts.references + i;
            if(ref->offset > text_len) {
                ref->offset = text_len;
            }
        }

        _hmml_persist_array(p, &p->out.timestamps, &p->out.timestamp_count, ts);
    }
}

static HMML_Credit _hmml_parse_credit(struct _hmml_parser* p, struct _hmml_str value)
{
    HMML_Credit credit = {};
    char* colon = strchr(value.ptr, ':');
    if(colon) {
        *colon = '\0';
        credit.name = _hmml_persist_str(p, (struct _hmml_str){ value.ptr, colon - value.ptr });
        credit.role = _hmml_persist_str(p, (struct _hmml_str){ colon+1, value.len - ((colon+1) - value.ptr) });
    } else {
        credit.name = _hmml_persist_str(p, value);
    }
    return credit;
}

static void _hmml_parse_video(struct _hmml_parser* p)
{
    struct str_attr {
        struct _hmml_str str;
        char** dest;
    } str_attrs[] = {
        { HSTR("stream_platform"), &p->out.metadata.stream_platform },
        { HSTR("project")        , &p->out.metadata.project },
        { HSTR("title")          , &p->out.metadata.title },
        { HSTR("vod_platform")   , &p->out.metadata.vod_platform },
        { HSTR("id")             , &p->out.metadata.id },
        { HSTR("template")       , &p->out.metadata.template },
        { HSTR("medium")         , &p->out.metadata.medium },
        { HSTR("number")         , &p->out.metadata.number },
        { HSTR("output")         , &p->out.metadata.output },
    };

    for(;;) {
next_attr:
        _hmml_skip_ws(p);

        if(*p->cursor == ']') {
            ++p->cursor;
            _hmml_parse_timestamps(p);
            return;
        }

        struct _hmml_str key, value;
        _hmml_read_kv(p, &key, &value);

        for(size_t i = 0; i < countof(str_attrs); ++i) {
            struct str_attr* s = str_attrs + i;
            if(_hmml_str_eq(key, s->str)) {
                *s->dest = _hmml_persist_str(p, value);
                goto next_attr;
            }
        }

        if(_hmml_str_eq(key, HSTR("credit"))) {
            HMML_Credit credit = _hmml_parse_credit(p, value);
            _hmml_persist_array(p, &p->out.metadata.credits, &p->out.metadata.credit_count, credit);
            goto next_attr;
        }

        if(_hmml_str_eq(key, HSTR("uncredit"))) {
            HMML_Credit uncredit = _hmml_parse_credit(p, value);
            _hmml_persist_array(p, &p->out.metadata.uncredits, &p->out.metadata.uncredit_count, uncredit);
            goto next_attr;
        }

        HMML_VideoCustomMetaData custom = {
            .key   = _hmml_persist_str(p, key),
            .value = _hmml_persist_str(p, value),
        };

        _hmml_persist_array(p, &p->out.metadata.custom, &p->out.metadata.custom_count, custom);
    }
}

HMML_Output hmml_parse(const char* string)
{
    struct _hmml_parser p = {
        .mem = string,
        .cursor = string,
        .line = 1,
    };

    if(setjmp(p.err_buf) == 1) {
        // if it returns 1, an error happened
        p.out.free_list = p.free_list;
        return p.out;
    }

    const struct _hmml_str prefix = HSTR("[video");
    if(strncmp(p.cursor, prefix.ptr, prefix.len)) {
        _hmml_err(&p, "Missing initial video tag.");
    } else {
        p.cursor += prefix.len;
        _hmml_parse_video(&p);
    }

    p.out.free_list = p.free_list;
    p.out.well_formed = 1;
    return p.out;
}

void hmml_free(HMML_Output* out)
{
    if(!out->free_list) {
        return;
    }

    for(uintptr_t i = 2; i < ((uintptr_t*)out->free_list)[1]; ++i) {
        free(((void**)out->free_list)[i]);
    }

    free(out->free_list);
}

const struct HMML_Version hmml_version = {
    2, 0, 13
};

#undef HSTX
#undef HSTR

#endif