Parsee/src/XML/SAX.c

#include <XML.h>

#include <Cytoplasm/HashMap.h>
#include <Cytoplasm/Memory.h>
#include <Cytoplasm/Array.h>
#include <Cytoplasm/Str.h>
#include <Cytoplasm/Log.h>

#include <string.h>
#include <stdlib.h>
#include <ctype.h>

struct XMLexer {
    Stream *stream;
    bool autofree;

    /* A temporary buffer */
    int *buffer;
    ssize_t length;
    ssize_t top;
   
    enum {
        XML_STATE_NONE              = 0,
        XML_STATE_COMMENT,

        XML_STATE_ATTR,
        XML_STATE_ATTRHEAD_PROP,
        XML_STATE_ATTRTAIL,
        XML_STATE_PI
    } state;

    struct {
        Array *elements;
        char *str;
    } data;
};

/* "Looks ahead" in the XML stream for a NUL-terminated string.
 * If it was found and skip is set, then also skips over it. */
static bool XMLookahead(XMLexer *lexer, const char *str, bool skip);

/* Parses an XML "name" */
static char * XMLParseName(XMLexer *lexer);
static bool XMLSkipSpace(XMLexer *lexer);
static char * XMLParseAttValue(XMLexer *lexer);

static ssize_t XMLInitialiseBuffer(XMLexer *lexer);
static int XMLGetc(XMLexer *lexer);
static void XMLUngetc(XMLexer *lexer, int ch);
static void XMLEndBuffer(XMLexer *lexer);
static void XMLRollback(XMLexer *lexer);
static void XMLReset(XMLexer *lexer, ssize_t loc);
static char * XMLStringify(XMLexer *lexer, ssize_t point);
static HashMap * XMLReadProps(XMLexer *l);
static void XMLPushElement(XMLexer *lexer, char *e);
static char * XMLPopElement(XMLexer *lexer);

static XMLEvent * XMLCreateEmptyElem(XMLexer *lexer, HashMap *attrs);
static XMLEvent * XMLCreateStart(XMLexer *lexer, HashMap *attrs);
static XMLEvent * XMLCreateRelax(XMLexer *lexer);
static XMLEvent * XMLCreateEnd(XMLexer *lexer, char *end);
static XMLEvent * XMLCreateData(XMLexer *lexer);

XMLexer *
XMLCreateLexer(Stream *stream, bool autofree)
{
    XMLexer *lexer;
    if (!stream)
    {
        return NULL;
    }

    lexer = Malloc(sizeof(*lexer));
    lexer->stream = stream;
    lexer->autofree = autofree;
    lexer->state = XML_STATE_NONE;
    
    lexer->buffer = NULL;
    lexer->length = -1;
    lexer->top = -1;

    lexer->data.elements = ArrayCreate();
    lexer->data.str = NULL;

    return lexer;
}
void
XMLFreeLexer(XMLexer *lexer)
{
    if (!lexer)
    {
        return;
    }

    if (lexer->autofree)
    {
        StreamClose(lexer->stream);
    }
    if (lexer->buffer)
    {
        Free(lexer->buffer);
    }
    if (lexer->data.str)
    {
        Free(lexer->data.str);
    }
    if (lexer->data.elements)
    {
        size_t i;
        Array *elems = lexer->data.elements;
        for (i = 0; i < ArraySize(elems); i++)
        {
            Free(ArrayGet(elems, i)); 
        }
        ArrayFree(lexer->data.elements);
    }
    Free(lexer);
}


XMLEvent *
XMLCrank(XMLexer *lexer)
{
    XMLEvent *event = NULL;
    char *attrname;
    HashMap *props;
    char cbuf[2] = { 0 };
    char *tmp;
    if (!lexer)
    {
        return NULL;
    }
    if (StreamEof(lexer->stream))
    {
        return NULL;
    }
    event = XMLCreateRelax(lexer);
    //Log(LOG_INFO, "A %d", lexer->state);

    switch (lexer->state)
    {
        case XML_STATE_NONE:
            if (XMLookahead(lexer, "<!--", true))
            {
                if (lexer->data.str)
                {
                    XMLFreeEvent(event);
                    event = XMLCreateData(lexer);
                }

                lexer->state = XML_STATE_COMMENT;
                break;
            }
            else if (XMLookahead(lexer, "</", true))
            {
                if (lexer->data.str)
                {
                    XMLFreeEvent(event);
                    event = XMLCreateData(lexer);
                }

                lexer->state = XML_STATE_ATTRTAIL;
                break;
            }
            else if (XMLookahead(lexer, "<?", true))
            {
                if (lexer->data.str)
                {
                    XMLFreeEvent(event);
                    event = XMLCreateData(lexer);
                }

                lexer->state = XML_STATE_PI;
                break;
            }
            else if (XMLookahead(lexer, "<", true))
            {
                if (lexer->data.str)
                {
                    XMLFreeEvent(event);
                    event = XMLCreateData(lexer);
                }

                lexer->state = XML_STATE_ATTR;
                break;
            }
            /* TODO: Try storing character into str buffer. */
            cbuf[0] = XMLGetc(lexer);
            tmp = lexer->data.str;
            lexer->data.str = StrConcat(2, tmp, cbuf);
            Free(tmp);
            break;
        case XML_STATE_COMMENT:
            if (XMLookahead(lexer, "-->", true))
            {
                lexer->state = XML_STATE_NONE;
            }
            else if (XMLookahead(lexer, "--", false))
            {
                /* Throw error */
                return NULL;
            }
            break;
        case XML_STATE_PI:
            if (XMLookahead(lexer, "?>", true))
            {
                lexer->state = XML_STATE_NONE;
                break;
            }
            XMLGetc(lexer);
            break;
        case XML_STATE_ATTR:
            attrname = XMLParseName(lexer);
            //Log(LOG_INFO, "A %d %s", lexer->state, attrname);
            if (!attrname)
            {
                /* TODO: Throw error */
            }
            XMLPushElement(lexer, attrname);

            //Log(LOG_INFO, "Reading props...");
            props = XMLReadProps(lexer);
            //Log(LOG_INFO, "Read props!");

            XMLSkipSpace(lexer);
            if (XMLookahead(lexer, "/>", true))
            {
                lexer->state = XML_STATE_NONE;
                XMLFreeEvent(event);
                event = XMLCreateEmptyElem(lexer, props);

                Free(XMLPopElement(lexer));
                break;
            }
            else if (XMLookahead(lexer, ">", true))
            {
                lexer->state = XML_STATE_NONE;
                XMLFreeEvent(event);
                event = XMLCreateStart(lexer, props);
                break;
            }
            else if (XMLookahead(lexer, "'", true))
            {
                while (true);
            }
            break;
        case XML_STATE_ATTRTAIL:
            attrname = XMLParseName(lexer);
            Free(XMLPopElement(lexer));
            if (!XMLookahead(lexer, ">", true))
            {
                /* TODO: Throw error. */
                break;
            }
            lexer->state = XML_STATE_NONE;
            XMLFreeEvent(event);
            event = XMLCreateEnd(lexer, attrname);
            break;
        default:
            /* TODO */
            break;
    }
    /* TODO: Crank our XML parser. */
    return event;
}
void
XMLFreeEvent(XMLEvent *event)
{
    if (!event)
    {
        return;
    }
    if (event->element)
    {
        Free(event->element);
    }
    if (event->attrs)
    {
        char *key;
        void *val;
        while (HashMapIterate(event->attrs, &key, &val))
        {
            Free(val);
        }
        HashMapFree(event->attrs);
    }
    if (event->data)
    {
        Free(event->data);
    }
    Free(event);
}

static bool
XMLookahead(XMLexer *lexer, const char *str, bool skip)
{
    int *stack;
    size_t top, i;
    ssize_t ntop;
    bool ret = false;
    if (!lexer || !str)
    {
        return false;
    }

    top = 0;
    stack = Malloc(strlen(str) * sizeof(*stack));
    
    for (i = 0; i < strlen(str); i++)
    {
        char c = str[i];
        int getc = XMLGetc(lexer);

        stack[top++] = getc;
        if (getc != c || getc == EOF)
        {
            goto seekback;
        }
    }

    /* We have been able to seek the string properly */
    ret = true;
    if (!skip)
    {
        goto seekback;
    }
    Free(stack);
    return ret;

seekback:
    for (ntop = top - 1; ntop >= 0; ntop--)
    {
        XMLUngetc(lexer, stack[ntop]);
    }
    Free(stack);
    return ret;
}

#define IsNamestart(c) ((c == ':') || isalpha(c) || (c == '_'))
#define IsNamepart(c) (IsNamestart(c) || (c == '-') || isdigit(c))
static char * 
XMLParseName(XMLexer *lexer)
{
    int c;
    ssize_t point;

    point = XMLInitialiseBuffer(lexer);
    c = XMLGetc(lexer);
    if (!IsNamestart(c))
    {
        XMLRollback(lexer);
        return NULL;
    }

    while ((c = XMLGetc(lexer)))
    {
        if (!IsNamepart(c))
        {
            break;
        }
    }
    XMLUngetc(lexer, c);

    return XMLStringify(lexer, point);
}

static ssize_t
XMLInitialiseBuffer(XMLexer *lexer)
{
    if (!lexer)
    {
        return -1;
    }
    if (lexer->length != -1)
    {
        return lexer->top;
    }

    lexer->length = 0;
    lexer->top = 0;
    lexer->buffer = NULL;
    return 0;
}
static int 
XMLGetc(XMLexer *lexer)
{
    int ch;
    if (lexer->length == -1)
    {
        return StreamGetc(lexer->stream);
    }

    ch = StreamGetc(lexer->stream);
    if (lexer->top >= lexer->length)
    {
        lexer->length += 8;
        lexer->buffer = Realloc(
            lexer->buffer, 
            lexer->length * sizeof(*lexer->buffer)
        );
    }
    lexer->buffer[lexer->top++] = ch;
    return ch;
}
static void 
XMLUngetc(XMLexer *lexer, int ch)
{
    if (!lexer || lexer->length == -1)
    {
        StreamUngetc(lexer->stream, ch);
        return;
    }
    if (lexer->top == 0)
    {
        return;
    }

    StreamUngetc(lexer->stream, ch);
    lexer->top--;
}
static void 
XMLEndBuffer(XMLexer *lexer)
{
    if (!lexer || lexer->length == -1)
    {
        return;
    }
    lexer->top = -1;
    lexer->length = -1;

    Free(lexer->buffer);
    lexer->buffer = NULL;
}
static void 
XMLRollback(XMLexer *lexer)
{
    ssize_t i;
    if (!lexer || lexer->length == -1)
    {
        return;
    }
    for (i = lexer->top - 1; i >= 0; i--)
    {
        StreamUngetc(lexer->stream, lexer->buffer[i]);
    }
    XMLEndBuffer(lexer);
}
void
XMLReset(XMLexer *lexer, ssize_t loc)
{
    ssize_t i;
    if (!lexer || lexer->length == -1)
    {
        return;
    }
    for (i = lexer->top - 1; i >= loc; i--)
    {
        StreamUngetc(lexer->stream, lexer->buffer[i]);
    }
    lexer->top = loc;
}
char *
XMLStringify(XMLexer *lexer, ssize_t point)
{
    ssize_t i;
    char *str;
    if (!lexer || lexer->length == -1)
    {
        return NULL;
    }
    str = Malloc(lexer->top + 1);
    memset(str, '\0', lexer->top + 1);
    for (i = point; i < lexer->top; i++)
    {
        str[i - point] = (char) lexer->buffer[i];
    }
    XMLEndBuffer(lexer);

    return str;
}

#define IsSpace(c) ((c == ' ') || (c == 0x09) || (c == 0x0D) || (c == 0x0A))
bool
XMLSkipSpace(XMLexer *lexer)
{
    int c;
    bool r = false;
    while ((c = XMLGetc(lexer)) != EOF)
    {
        if (!IsSpace(c))
        {
            break;
        }
        r = true;
    }
    XMLUngetc(lexer, c);

    return r;
}
HashMap *
XMLReadProps(XMLexer *lexer)
{
    ssize_t point = XMLInitialiseBuffer(lexer);
    HashMap *map = NULL;
    bool error = false;
    while (true)
    {
        char *name;
        char *value = NULL;

        if (!XMLSkipSpace(lexer))
        {
            /* A lack of space is totally excepted */
            break;
        }

        name = XMLParseName(lexer);
        if (!name)
        {
            /* A lack of name is totally excepted */
            break;
        }
        //Log(LOG_INFO, "K=%s...", name);

        value = StrDuplicate("");
        if (XMLookahead(lexer, "=", true))
        {
            /* TODO: Values aren't names. */
            Free(value);
            value = XMLParseAttValue(lexer);
            if (!value)
            {
                error = true;
                Free(name);
                break;
            }
        }
        if (!map)
        {
            map = HashMapCreate();
        }
        HashMapSet(map, name, value);
        Free(name);
    }
    if (error || !map)
    {
        XMLReset(lexer, point);
    }
    return map;
}
static XMLEvent * 
XMLCreateStart(XMLexer *lexer, HashMap *attrs)
{
    XMLEvent *event = Malloc(sizeof(*event));
    size_t elements = ArraySize(lexer->data.elements) - 1;
    char *h_element = StrDuplicate(ArrayGet(lexer->data.elements, elements));

    event->type = XML_LEXER_STARTELEM;
    event->element = h_element;
    event->attrs = attrs;
    event->data = NULL;

    /* TODO */
    event->line = 0;
    event->col = 0;
    event->offset = 0;

    return event;
}
XMLEvent *
XMLCreateEnd(XMLexer *lexer, char *end)
{
    XMLEvent *event = Malloc(sizeof(*event));

    event->type = XML_LEXER_ENDELEM;
    event->element = end;
    event->attrs = NULL;
    event->data = NULL;

    /* TODO */
    event->line = 0;
    event->col = 0;
    event->offset = 0;

    (void) lexer;

    return event;
}
static XMLEvent *
XMLCreateEmptyElem(XMLexer *lexer, HashMap *attrs)
{
    XMLEvent *event = Malloc(sizeof(*event));
    size_t elements = ArraySize(lexer->data.elements) - 1;
    char *h_element = StrDuplicate(ArrayGet(lexer->data.elements, elements));

    event->type = XML_LEXER_ELEM;
    event->element = h_element;
    event->attrs = attrs;
    event->data = NULL;

    /* TODO */
    event->line = 0;
    event->col = 0;
    event->offset = 0;

    return event;
}

static char *
XMLDecodeString(char *s)
{
    char *ret = NULL, *tmp;
    char cs[2] = { 0 };
    
    while (*s)
    {
        cs[0] = *s; 
        if (!strncmp(s, "&apos;", 6))
        {
            cs[0] = '\'';
            s += 6;
        }
        else if (!strncmp(s, "&quot;", 6))
        {
            cs[0] = '"';
            s += 6;
        }
        else if (!strncmp(s, "&lt;", 4))
        {
            cs[0] = '<';
            s += 4;
        }
        else if (!strncmp(s, "&gt;", 4))
        {
            cs[0] = '>';
            s += 4;
        }
        else if (!strncmp(s, "&amp;", 5))
        {
            cs[0] = '&';
            s += 5;
        }
        else if (!strncmp(s, "&#", 2))
        {
            char *dec = s + 2;
            char *end = strchr(dec, ';');
            if (!end)
            {
                s++;
            }
            else
            {
                /* TODO: Decode any Unicode glyph as UTF-8. */
                int val = strtol(dec, &end, 10);
                cs[0] = val;
                s = end + 1;
            }
        }
        else
        {
            s++;
        }
        /* TODO: Support hexcodes */

        tmp = ret;
        ret = StrConcat(2, ret, cs);
        Free(tmp);
    }
    return ret;
}
XMLEvent *
XMLCreateData(XMLexer *lexer)
{
    XMLEvent *event = Malloc(sizeof(*event));
    size_t elements = ArraySize(lexer->data.elements);

    event->type = XML_LEXER_DATA;
    event->element = elements ?
        StrDuplicate(ArrayGet(lexer->data.elements, elements - 1)) :
        NULL;
    event->attrs = NULL;
    event->data = XMLDecodeString(lexer->data.str);

    /* TODO */
    event->line = 0;
    event->col = 0;
    event->offset = 0;

    Free(lexer->data.str);
    lexer->data.str = NULL;

    return event;
}
static XMLEvent * 
XMLCreateRelax(XMLexer *lexer)
{
    XMLEvent *event = Malloc(sizeof(*event));
    size_t elements = ArraySize(lexer->data.elements);

    event->type = XML_RELAX;
    event->element = elements ?
        StrDuplicate(ArrayGet(lexer->data.elements, elements - 1)) :
        NULL;
    event->attrs = NULL;
    event->data = NULL;

    /* TODO */
    event->line = 0;
    event->col = 0;
    event->offset = 0;

    return event;
}

static void 
XMLPushElement(XMLexer *lexer, char *e)
{
    ArrayAdd(lexer->data.elements, e);
}
static char * 
XMLPopElement(XMLexer *lexer)
{
    size_t n = ArraySize(lexer->data.elements);
    if (n == 0)
    {
        return NULL;
    }

    return ArrayDelete(lexer->data.elements, n - 1);
}

#define IsNormalQ(c) ((c != '<') && (c != '&') && (c != '\''))
#define IsNormalD(c) ((c != '<') && (c != '&') && (c != '"'))
static char *
XMLParseAttQuote(XMLexer *lexer)
{
    int c;
    ssize_t point;
    char *str;

    point = XMLInitialiseBuffer(lexer);

    while ((c = XMLGetc(lexer)))
    {
        if (c == '&')
        {
            //char *code = NULL;
            int c2;
            int p2 = XMLInitialiseBuffer(lexer);
            int j = 0;
            while ((c2 = XMLGetc(lexer)) && c2 != EOF && j < 8)
            {
                if (c2 == ';')
                {
                    break;
                }
                j++;
            }
            if (c2 != ';')
            {
                XMLReset(lexer, p2);
            }
        }
        else if (!IsNormalQ(c))
        {
            break;
        }
    }
    if (c != '\'')
    {
        XMLUngetc(lexer, c);
        return NULL;
    }

    /* TODO: Decode the string */
    str = XMLStringify(lexer, point);
    str[strlen(str) - 1] = '\0'; /* Trim the quote. */
    return str;
}
static char *
XMLParseAttDouble(XMLexer *lexer)
{
    int c;
    ssize_t point;
    char *str;

    point = XMLInitialiseBuffer(lexer);

    while ((c = XMLGetc(lexer)))
    {
        //Log(LOG_INFO, "E2=%c", c);
        if (c == '&')
        {
            int c2;
            int p2 = XMLInitialiseBuffer(lexer);
            int j = 0;
            while ((c2 = XMLGetc(lexer)) && c2 != EOF && j < 8)
            {
                if (c2 == ';')
                {
                    break;
                }
                j++;
            }
            if (c2 != ';')
            {
                XMLReset(lexer, p2);
            }
            continue;
        }
        else if (!IsNormalD(c))
        {
            break;
        }
    }
    if (c != '"')
    {
        XMLUngetc(lexer, c);
        return NULL;
    }

    /* TODO: Decode the string */
    str = XMLStringify(lexer, point);
    str[strlen(str) - 1] = '\0'; /* Trim the quote. */
    return str;
}
static char *
XMLParseAttValue(XMLexer *lexer)
{
    XMLInitialiseBuffer(lexer);

    if (XMLookahead(lexer, "'", true))
    {
        return XMLParseAttQuote(lexer);
    }
    else if (XMLookahead(lexer, "\"", true))
    {
        return XMLParseAttDouble(lexer);
    }
    return NULL;
}