src/json.c

/*
 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS HEADER.
 *
 * Copyright 2024 Mike Becker, Olaf Wintermann All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions are met:
 *
 *   1. Redistributions of source code must retain the above copyright
 *      notice, this list of conditions and the following disclaimer.
 *
 *   2. Redistributions in binary form must reproduce the above copyright
 *      notice, this list of conditions and the following disclaimer in the
 *      documentation and/or other materials provided with the distribution.
 *
 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
 * POSSIBILITY OF SUCH DAMAGE.
 */

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

#include "cx/json.h"

/*
 * RFC 8259
 * https://tools.ietf.org/html/rfc8259
 */

#define PARSER_READVALUE_ALLOC 32

static CxJsonValue cx_json_value_nothing = {.type = CX_JSON_NOTHING};

static void token_destroy(CxJsonToken *token) {
    if (token->alloc > 0) {
        free((char*) token->content);
    }
}

static int token_append(CxJsonToken *token, const char *buf, size_t len) {
    if (len == 0) {
        return 0;
    }

    size_t newlen = token->length + len;
    if (token->alloc < newlen) {
        char *newbuf = realloc(
                token->alloc == 0 ? NULL : (char *) token->content,
                newlen);
        if (!newbuf) {
            return 1;
        }
        token->content = newbuf;
        token->alloc = newlen;
    }

    memcpy((char *) token->content + token->length, buf, len);
    token->length = newlen;
    return 0;
}

static CxJsonToken token_create(CxJson *json, size_t start, size_t end) {
    CxJsonToken token = {0};
    size_t len = end - start;
    if (json->uncompleted.tokentype == CX_JSON_NO_TOKEN) {
        token.content = json->buffer + start;
        token.length = len;
    } else {
        if (token_append(&json->uncompleted, json->buffer + start, len)) {
            // TODO: this does certainly not lead to correct error handling
            return (CxJsonToken){0};
        }
        token = json->uncompleted;
    }
    json->uncompleted = (CxJsonToken){0};
    return token;
}

static int token_isliteral(const char *content, size_t length) {
    if (length == 4) {
        if (!memcmp(content, "true", 4)) {
            return 1;
        } else if (!memcmp(content, "null", 4)) {
            return 1;
        }
    } else if (length == 5 && !memcmp(content, "false", 5)) {
        return 1;
    }
    return 0;
}

static int num_isexp(const char *content, size_t length, size_t pos) {
    if (pos >= length) {
        return 0;
    }

    int ok = 0;
    for (size_t i = pos; i < length; i++) {
        char c = content[i];
        if (isdigit(c)) {
            ok = 1;
        } else if (i == pos) {
            if (!(c == '+' || c == '-')) {
                return 0;
            }
        } else {
            return 0;
        }
    }

    return ok;
}

static CxJsonTokenType token_numbertype(const char *content, size_t length) {
    if (length == 0) return CX_JSON_TOKEN_ERROR;

    if (content[0] != '-' && !isdigit(content[0])) {
        return CX_JSON_TOKEN_ERROR;
    }

    CxJsonTokenType type = CX_JSON_TOKEN_INTEGER;
    for (size_t i = 1; i < length; i++) {
        if (content[i] == '.') {
            if (type == CX_JSON_TOKEN_NUMBER) {
                return CX_JSON_TOKEN_ERROR; // more than one decimal separator
            }
            type = CX_JSON_TOKEN_NUMBER;
        } else if (content[i] == 'e' || content[i] == 'E') {
            return num_isexp(content, length, i + 1) ? CX_JSON_TOKEN_NUMBER : CX_JSON_TOKEN_ERROR;
        } else if (!isdigit(content[i])) {
            return CX_JSON_TOKEN_ERROR; // char is not a digit, decimal separator or exponent sep
        }
    }

    return type;
}

static CxJsonTokenType char2ttype(char c) {
    switch (c) {
        case '[': {
            return CX_JSON_TOKEN_BEGIN_ARRAY;
        }
        case '{': {
            return CX_JSON_TOKEN_BEGIN_OBJECT;
        }
        case ']': {
            return CX_JSON_TOKEN_END_ARRAY;
        }
        case '}': {
            return CX_JSON_TOKEN_END_OBJECT;
        }
        case ':': {
            return CX_JSON_TOKEN_NAME_SEPARATOR;
        }
        case ',': {
            return CX_JSON_TOKEN_VALUE_SEPARATOR;
        }
        case '"': {
            return CX_JSON_TOKEN_STRING;
        }
        default: {
            if (isspace(c)) {
                return CX_JSON_TOKEN_SPACE;
            }
        }
    }
    return CX_JSON_NO_TOKEN;
}

static CxJsonToken token_parse_next(CxJson *json) {
    // current token type and start index
    CxJsonTokenType ttype = json->uncompleted.tokentype;
    size_t token_start = json->pos;

    for (size_t i = json->pos; i < json->size; i++) {
        char c = json->buffer[i];
        if (ttype != CX_JSON_TOKEN_STRING) {
            // currently non-string token

            CxJsonTokenType ctype = char2ttype(c); // start of new token?

            if (ttype == CX_JSON_NO_TOKEN) {
                if (ctype == CX_JSON_TOKEN_SPACE) {
                    continue;
                } else if (ctype == CX_JSON_TOKEN_STRING) {
                    // begin string
                    ttype = CX_JSON_TOKEN_STRING;
                    token_start = i;
                } else if (ctype != CX_JSON_NO_TOKEN) {
                    // single-char token
                    json->pos = i + 1;
                    CxJsonToken token = {ctype, NULL, 0, 0};
                    return token;
                } else {
                    ttype = CX_JSON_TOKEN_LITERAL; // number or literal
                    token_start = i;
                }
            } else {
                // finish token
                if (ctype != CX_JSON_NO_TOKEN) {
                    CxJsonToken ret = token_create(json, token_start, i);
                    if (token_isliteral(ret.content, ret.length)) {
                        ret.tokentype = CX_JSON_TOKEN_LITERAL;
                    } else {
                        ret.tokentype = token_numbertype(ret.content, ret.length);
                    }
                    json->pos = i;
                    return ret;
                }
            }
        } else {
            // currently inside a string
            if (json->tokenizer_escape) {
                json->tokenizer_escape = false;
            } else {
                if (c == '"') {
                    CxJsonToken ret = token_create(json, token_start, i + 1);
                    ret.tokentype = CX_JSON_TOKEN_STRING;
                    json->pos = i + 1;
                    return ret;
                } else if (c == '\\') {
                    json->tokenizer_escape = true;
                }
            }
        }
    }

    if (ttype != CX_JSON_NO_TOKEN) {
        // uncompleted token
        size_t uncompeted_len = json->size - token_start;
        if (json->uncompleted.tokentype == CX_JSON_NO_TOKEN) {
            // current token is uncompleted
            // save current token content in p->uncompleted
            CxJsonToken uncompleted;
            uncompleted.tokentype = ttype;
            uncompleted.length = uncompeted_len;
            uncompleted.alloc = uncompeted_len + 16;
            char *tmp = malloc(uncompleted.alloc);
            if (tmp) {
                memcpy(tmp, json->buffer + token_start, uncompeted_len);
                uncompleted.content = tmp;
                json->uncompleted = uncompleted;
            } else {
                json->error = 1;
            }
        } else {
            // previously we also had an uncompleted token
            // combine the uncompleted token with the current token
            if (token_append(&json->uncompleted, json->buffer + token_start, uncompeted_len)) {
                json->error = 1;
            }
        }
    }

    CxJsonToken ret = {CX_JSON_NO_TOKEN, NULL, 0, 0};
    return ret;
}

static cxmutstr unescape_string(const CxAllocator *a, const char *str, size_t len) {
    // TODO: support more escape sequences
    // we know that the unescaped string will be shorter by at least 2 chars
    cxmutstr result;
    result.length = 0;
    result.ptr = cxMalloc(a, len - 1);
    if (result.ptr == NULL) {
        return result;
    }

    bool u = false;
    for (size_t i = 1; i < len - 1; i++) {
        char c = str[i];
        if (u) {
            u = false;
            if (c == 'n') {
                c = '\n';
            } else if (c == 't') {
                c = '\t';
            }
            result.ptr[result.length++] = c;
        } else {
            if (c == '\\') {
                u = true;
            } else {
                result.ptr[result.length++] = c;
            }
        }
    }
    result.ptr[result.length] = 0;

    return result;
}

static int parse_number(const char *str, size_t len, void *value, bool asint) {
    char *endptr = NULL;
    char buf[32];
    if (len > 30) {
        return 1;
    }
    // TODO: if we can guarantee that we are working on a copied string already, we can avoid this memcpy
    memcpy(buf, str, len);
    buf[len] = 0;

    if (asint) {
        long long v = strtoll(buf, &endptr, 10);
        *((int64_t*)value) = (int64_t) v;
    } else {
        // TODO: proper JSON spec number parser
        double v = strtod(buf, &endptr);
        *((double*)value) = v;
    }

    return (endptr != &buf[len]);
}

static CxJsonValue* create_json_value(CxJson *json, CxJsonValueType type) {
    CxJsonValue *v = cxMalloc(json->allocator, sizeof(CxJsonValue));
    if (v == NULL) {
        return NULL;
    }

    // initialize the value
    if (type == CX_JSON_ARRAY) {
        cx_array_initialize_a(json->allocator, v->value.array.array, 16);
        if (v->value.array.array == NULL) {
            cxFree(json->allocator, v);
            return NULL;
        }
    } else if (type == CX_JSON_OBJECT) {
        cx_array_initialize_a(json->allocator, v->value.object.values, 16);
        if (v->value.object.values == NULL) {
            cxFree(json->allocator, v);
            return NULL;
        }
    } else {
        memset(v, 0, sizeof(CxJsonValue));
    }
    v->type = type;
    v->allocator = json->allocator;

    // add the new value to a possible parent
    CxArrayReallocator value_realloc = cx_array_reallocator(json->allocator, NULL);
    if (json->vbuf_size > 0) {
        CxJsonValue *parent = json->vbuf[json->vbuf_size - 1];
        if (parent->type == CX_JSON_ARRAY) {
            cx_array_simple_add_a(&value_realloc, parent->value.array.array, v);
        } else if (parent->type == CX_JSON_OBJECT) {
            assert(parent->value.object.values_size > 0);
            assert(parent->value.object.values[parent->value.object.values_size - 1].value == NULL);
            parent->value.object.values[parent->value.object.values_size - 1].value = v;
        } else {
            assert(false);
        }
    }

    // add the new value to the stack, if it is an array or object
    if (type == CX_JSON_ARRAY || type == CX_JSON_OBJECT) {
        CxArrayReallocator vbuf_realloc = cx_array_reallocator(NULL, json->vbuf_internal);
        if (cx_array_simple_add_a(&vbuf_realloc, json->vbuf, v)) {
            cxFree(json->allocator, v);
            return NULL;
        }
    }

    // if currently no value is parsed, this is now the value of interest
    if (json->parsed == NULL) {
        json->parsed = v;
    }

    return v;
}

static int json_obj_add_entry(CxJson *json, char *name) {
    CxJsonObjValue kv = {name, NULL};
    assert(json->vbuf_size > 0);
    CxJsonValue *parent = json->vbuf[json->vbuf_size - 1];
    assert(parent != NULL);
    assert(parent->type == CX_JSON_OBJECT);
    CxArrayReallocator value_realloc = cx_array_reallocator(json->allocator, NULL);
    return cx_array_simple_add_a(&value_realloc, parent->value.object.values, kv);
}

#define JP_STATE_VALUE_BEGIN         0
#define JP_STATE_VALUE_END          10
#define JP_STATE_VALUE_BEGIN_OBJ     1
#define JP_STATE_OBJ_SEP_OR_CLOSE   11
#define JP_STATE_VALUE_BEGIN_AR      2
#define JP_STATE_ARRAY_SEP_OR_CLOSE 12
#define JP_STATE_OBJ_NAME_OR_CLOSE   5
#define JP_STATE_OBJ_NAME            6
#define JP_STATE_OBJ_COLON           7

void cxJsonInit(CxJson *json, const CxAllocator *allocator) {
    if (allocator == NULL) {
        allocator = cxDefaultAllocator;
    }
    
    memset(json, 0, sizeof(CxJson));
    json->allocator = allocator;

    json->states = json->states_internal;
    json->states_capacity = cx_nmemb(json->states_internal);
    json->states[0] = JP_STATE_VALUE_BEGIN;
    json->states_size = 1;

    json->vbuf = json->vbuf_internal;
    json->vbuf_capacity = cx_nmemb(json->vbuf_internal);
}

void cxJsonDestroy(CxJson *json) {
    if (json->states != json->states_internal) {
        free(json->states);
    }
    if (json->vbuf != json->vbuf_internal) {
        free(json->vbuf);
    }
    cxJsonValueFree(json->parsed);
    json->parsed = NULL;
}

int cxJsonFilln(CxJson *json, const char *buf, size_t size) {
    // TODO: implement rescue buffer like in CxProperties to allow subsequent fills
    json->buffer = buf;
    json->size = size;
    json->pos = 0;
    return 0;
}

static void json_add_state(CxJson *json, int state) {
    // we have guaranteed the necessary space with cx_array_simple_reserve()
    // therefore, we can safely add the state in the simplest way possible
    json->states[json->states_size++] = state;
}

#define return_rec(code) \
    token_destroy(&token); \
    return code

static int json_parse(CxJson *json) {
    // Reserve a pointer for a possibly read value
    CxJsonValue *vbuf = NULL;

    // grab the next token
    CxJsonToken token = token_parse_next(json);
    if (token.tokentype == CX_JSON_NO_TOKEN) {
        // nothing found, wait for more data
        return 0;
    }

    // pop the current state
    assert(json->states_size > 0);
    int state = json->states[--json->states_size];

    // guarantee that at least two more states fit on the stack
    CxArrayReallocator state_realloc = cx_array_reallocator(NULL, json->states_internal);
    if (cx_array_simple_reserve_a(&state_realloc, json->states, 2)) {
        return -1;
    }


    //  0 JP_STATE_VALUE_BEGIN          value begin
    // 10 JP_STATE_VALUE_END            expect value end
    //  1 JP_STATE_VALUE_BEGIN_OBJ      value begin (inside object)
    // 11 JP_STATE_OBJ_SEP_OR_CLOSE     object, expect separator, objclose
    //  2 JP_STATE_VALUE_BEGIN_AR       value begin (inside array)
    // 12 JP_STATE_ARRAY_SEP_OR_CLOSE   array, expect separator or arrayclose
    //  5 JP_STATE_OBJ_NAME_OR_CLOSE    object, expect name or objclose
    //  6 JP_STATE_OBJ_NAME             object, expect name
    //  7 JP_STATE_OBJ_COLON            object, expect ':'

    if (state < 3) {
        // push expected end state to the stack
        json_add_state(json, 10 + state);
        switch (token.tokentype) {
            case CX_JSON_TOKEN_BEGIN_ARRAY: {
                if (create_json_value(json, CX_JSON_ARRAY) == NULL) {
                    // TODO: error code - no memory
                    return_rec(-1);
                }
                json_add_state(json, JP_STATE_VALUE_BEGIN_AR);
                return_rec(1);
            }
            case CX_JSON_TOKEN_BEGIN_OBJECT: {
                if (create_json_value(json, CX_JSON_OBJECT) == NULL) {
                    // TODO: error code - no memory
                    return_rec(-1);
                }
                json_add_state(json, JP_STATE_OBJ_NAME_OR_CLOSE);
                return_rec(1);
            }
            case CX_JSON_TOKEN_STRING: {
                if ((vbuf = create_json_value(json, CX_JSON_STRING)) == NULL) {
                    // TODO: error code - no memory
                    return_rec(-1);
                }
                cxmutstr str = unescape_string(json->allocator, token.content, token.length);
                if (str.ptr == NULL) {
                    // TODO: error code - no memory
                    return_rec(-1);
                }
                vbuf->value.string = str;
                return_rec(1);
            }
            case CX_JSON_TOKEN_INTEGER:
            case CX_JSON_TOKEN_NUMBER: {
                int type = token.tokentype == CX_JSON_TOKEN_INTEGER ? CX_JSON_INTEGER : CX_JSON_NUMBER;
                if (NULL == (vbuf = create_json_value(json, type))) {
                    // TODO: error code - no memory
                    return_rec(-1);
                }
                if (parse_number(token.content, token.length, &vbuf->value,type == CX_JSON_INTEGER)) {
                    // TODO: error code - format error
                    return_rec(-1);
                }
                return_rec(1);
            }
            case CX_JSON_TOKEN_LITERAL: {
                if ((vbuf = create_json_value(json, CX_JSON_LITERAL)) == NULL) {
                    // TODO: error code - no memory
                    return_rec(-1);
                }
                const char *l = token.content;
                size_t token_len = token.length;
                if (token_len == 4 && !memcmp(l, "true", 4)) {
                    vbuf->value.literal = CX_JSON_TRUE;
                } else if (token_len == 5 && !memcmp(l, "false", 5)) {
                    vbuf->value.literal = CX_JSON_FALSE;
                } else {
                    vbuf->value.literal = CX_JSON_NULL;
                }
                return_rec(1);
            }
            default: {
                // TODO: error code - unexpected token
                return_rec(-1);
            }
        }
    } else if (state == JP_STATE_ARRAY_SEP_OR_CLOSE) {
        // expect ',' or ']'
        if (token.tokentype == CX_JSON_TOKEN_VALUE_SEPARATOR) {
            json_add_state(json, JP_STATE_VALUE_BEGIN_AR);
            return_rec(1);
        } else if (token.tokentype == CX_JSON_TOKEN_END_ARRAY) {
            // discard the array from the value buffer
            json->vbuf_size--;
            return_rec(1);
        } else {
            // TODO: error code - unexpected token
            return_rec(-1);
        }
    } else if (state == JP_STATE_OBJ_NAME_OR_CLOSE || state == JP_STATE_OBJ_NAME) {
        if (state == JP_STATE_OBJ_NAME_OR_CLOSE && token.tokentype == CX_JSON_TOKEN_END_OBJECT) {
            // discard the obj from the value buffer
            json->vbuf_size--;
            return_rec(1);
        } else {
            // expect string
            if (token.tokentype != CX_JSON_TOKEN_STRING) {
                // TODO: error code - unexpected token
                return_rec(-1);
            }

            // add new entry
            cxmutstr name = unescape_string(json->allocator, token.content, token.length);
            if (name.ptr == NULL) {
                // TODO: error code - no mem
                return_rec(-1);
            }
            json_obj_add_entry(json, name.ptr);

            // next state
            json_add_state(json, JP_STATE_OBJ_COLON);
            return_rec(1);
        }
    } else if (state == JP_STATE_OBJ_COLON) {
        // expect ':'
        if (token.tokentype != CX_JSON_TOKEN_NAME_SEPARATOR) {
            // TODO: error code - unexpected token
            return_rec(-1);
        }
        // next state
        json_add_state(json, JP_STATE_VALUE_BEGIN_OBJ);
        return_rec(1);
    } else if (state == JP_STATE_OBJ_SEP_OR_CLOSE) {
        // expect ',' or '}'
        if (token.tokentype == CX_JSON_TOKEN_VALUE_SEPARATOR) {
            json_add_state(json, JP_STATE_OBJ_NAME);
            return_rec(1);
        } else if (token.tokentype == CX_JSON_TOKEN_END_OBJECT) {
            // discard the obj from the value buffer
            json->vbuf_size--;
            return_rec(1);
        } else {
            // TODO: error code - unexpected token
            return_rec(-1);
        }
    } else {
        // should be unreachable
        assert(false);
        return_rec(-1);
    }
}

int cxJsonNext(CxJson *json, CxJsonValue **value) {
    // TODO: replace int with a status enum like in CxProperties

    // initialize output value
    *value = &cx_json_value_nothing;

    // parse data
    int result;
    do {
        result = json_parse(json);
        if (result == 1 && json->states_size == 1) {
            // final state reached
            assert(json->states[0] == JP_STATE_VALUE_END);
            assert(json->vbuf_size == 0);

            // write output value
            *value = json->parsed;
            json->parsed = NULL;

            // re-initialize state machine
            json->states[0] = JP_STATE_VALUE_BEGIN;

            return 1;
        }
    } while (result == 1);

    return result;
}

void cxJsonValueFree(CxJsonValue *value) {
    if (value == NULL || value == &cx_json_value_nothing) return;

    switch (value->type) {
        case CX_JSON_OBJECT: {
            CxJsonObject obj = value->value.object;
            for (size_t i = 0; i < obj.values_size; i++) {
                cxJsonValueFree(obj.values[i].value);
                cxFree(value->allocator, obj.values[i].name);
            }
            cxFree(value->allocator, obj.values);
            break;
        }
        case CX_JSON_ARRAY: {
            CxJsonArray array = value->value.array;
            for (size_t i = 0; i < array.array_size; i++) {
                cxJsonValueFree(array.array[i]);
            }
            cxFree(value->allocator, array.array);
            break;
        }
        case CX_JSON_STRING: {
            cxFree(value->allocator, value->value.string.ptr);
            break;
        }
        default: {
            break;
        }
    }
    cxFree(value->allocator, value);
}

CxJsonValue *cxJsonArrGet(CxJsonValue *value, size_t index) {
    if (index >= value->value.array.array_size) {
        return &cx_json_value_nothing;
    }
    return value->value.array.array[index];
}

CxJsonValue *cxJsonObjGet(CxJsonValue *value, const char *name) {
    CxJsonObject *obj = &(value->value.object);
    // TODO: think about sorting the object so that we can use binary search here
    for (size_t i = 0; i < obj->values_size; i++) {
        // TODO: we might want to store names as cxmutstr
        if (0 == strcmp(name, obj->values[i].name)) {
            return obj->values[i].value;
        }
    }
    return &cx_json_value_nothing;
}