readstat_error_t readstat_parse_dta(readstat_parser_t *parser, const char *filename, void *user_ctx) {
readstat_error_t retval = READSTAT_OK;
int i;
size_t record_len = 0;
int fd = -1;
char *buf = NULL;
dta_header_t header;
dta_ctx_t *ctx = NULL;
char str_buf[2048];
char *long_string = NULL;
size_t file_size = 0;
if ((fd = readstat_open(filename)) == -1) {
retval = READSTAT_ERROR_OPEN;
goto cleanup;
}
char magic[4];
if (read(fd, magic, 4) != 4) {
retval = READSTAT_ERROR_READ;
goto cleanup;
}
file_size = readstat_lseek(fd, 0, SEEK_END);
if (file_size == -1) {
retval = READSTAT_ERROR_SEEK;
goto cleanup;
}
if (readstat_lseek(fd, 0, SEEK_SET) == -1) {
retval = READSTAT_ERROR_SEEK;
goto cleanup;
}
if (strncmp(magic, "<sta", 4) == 0) {
retval = dta_read_xmlish_preamble(fd, ctx, &header);
} else {
if (read(fd, &header, sizeof(header)) != sizeof(header)) {
retval = READSTAT_ERROR_READ;
goto cleanup;
}
}
if ((ctx = dta_ctx_init(header.nvar, header.nobs, header.byteorder, header.ds_format)) == NULL) {
retval = READSTAT_ERROR_MALLOC;
goto cleanup;
}
ctx->user_ctx = user_ctx;
ctx->file_size = file_size;
ctx->progress_handler = parser->progress_handler;
retval = dta_update_progress(fd, ctx);
if (retval != READSTAT_OK)
goto cleanup;
if (parser->info_handler) {
if (parser->info_handler(ctx->nobs, ctx->nvar, user_ctx)) {
retval = READSTAT_ERROR_USER_ABORT;
goto cleanup;
}
}
if (ctx->file_is_xmlish) {
uint16_t label_len = 0;
unsigned char timestamp_len;
if ((retval = dta_read_tag(fd, ctx, "<label>")) != READSTAT_OK) {
goto cleanup;
}
if (ctx->data_label_len_len == 2) {
if (read(fd, &label_len, sizeof(uint16_t)) != sizeof(uint16_t)) {
retval = READSTAT_ERROR_READ;
goto cleanup;
}
label_len = ctx->machine_needs_byte_swap ? byteswap2(label_len) : label_len;
} else if (ctx->data_label_len_len == 1) {
unsigned char label_len_char;
if (read(fd, &label_len_char, sizeof(unsigned char)) != sizeof(unsigned char)) {
retval = READSTAT_ERROR_READ;
goto cleanup;
}
label_len = label_len_char;
}
if (readstat_lseek(fd, label_len, SEEK_CUR) == -1) {
retval = READSTAT_ERROR_SEEK;
goto cleanup;
}
if ((retval = dta_read_tag(fd, ctx, "</label>")) != READSTAT_OK) {
goto cleanup;
}
if ((retval = dta_read_tag(fd, ctx, "<timestamp>")) != READSTAT_OK) {
goto cleanup;
}
if (read(fd, ×tamp_len, 1) != 1) {
retval = READSTAT_ERROR_READ;
goto cleanup;
}
if (readstat_lseek(fd, timestamp_len, SEEK_CUR) == -1) {
retval = READSTAT_ERROR_SEEK;
goto cleanup;
}
if ((retval = dta_read_tag(fd, ctx, "</timestamp>")) != READSTAT_OK) {
goto cleanup;
}
} else {
if (readstat_lseek(fd, ctx->data_label_len, SEEK_CUR) == -1) {
retval = READSTAT_ERROR_SEEK;
goto cleanup;
}
if (ctx->time_stamp_len) {
if (readstat_lseek(fd, ctx->time_stamp_len, SEEK_CUR) == -1) {
retval = READSTAT_ERROR_SEEK;
goto cleanup;
}
}
}
if ((retval = dta_read_tag(fd, ctx, "</header>")) != READSTAT_OK) {
goto cleanup;
}
if (dta_read_map(fd, ctx) != READSTAT_OK) {
retval = READSTAT_ERROR_READ;
goto cleanup;
}
if (dta_read_descriptors(fd, ctx) != READSTAT_OK) {
retval = READSTAT_ERROR_READ;
goto cleanup;
}
for (i=0; i<ctx->nvar; i++) {
size_t max_len;
readstat_types_t type = dta_type_info(ctx->typlist[i], &max_len, ctx);
record_len += max_len;
if (type == READSTAT_TYPE_STRING)
max_len++; /* might append NULL */
if (parser->variable_handler) {
readstat_variable_t *variable = dta_init_variable(ctx, i, type);
const char *value_labels = NULL;
if (ctx->lbllist[ctx->lbllist_entry_len*i])
value_labels = &ctx->lbllist[ctx->lbllist_entry_len*i];
int cb_retval = parser->variable_handler(i, variable, value_labels, user_ctx);
free(variable);
if (cb_retval) {
retval = READSTAT_ERROR_USER_ABORT;
goto cleanup;
}
}
}
if ((retval = dta_skip_expansion_fields(fd, ctx)) != READSTAT_OK) {
goto cleanup;
}
if (record_len == 0) {
retval = READSTAT_ERROR_PARSE;
goto cleanup;
}
if ((retval = dta_read_tag(fd, ctx, "<data>")) != READSTAT_OK) {
goto cleanup;
}
if ((retval = dta_update_progress(fd, ctx)) != READSTAT_OK) {
goto cleanup;
}
if ((buf = malloc(record_len)) == NULL) {
retval = READSTAT_ERROR_MALLOC;
goto cleanup;
}
for (i=0; i<ctx->nobs; i++) {
if (read(fd, buf, record_len) != record_len) {
retval = READSTAT_ERROR_READ;
goto cleanup;
}
int j;
off_t offset = 0;
for (j=0; j<ctx->nvar; j++) {
size_t max_len;
readstat_value_t value;
memset(&value, 0, sizeof(readstat_value_t));
value.type = dta_type_info(ctx->typlist[j], &max_len, ctx);
if (value.type == READSTAT_TYPE_STRING) {
readstat_convert(str_buf, sizeof(str_buf), &buf[offset], max_len, ctx->converter);
value.v.string_value = str_buf;
} else if (value.type == READSTAT_TYPE_LONG_STRING) {
uint32_t v, o;
v = *((uint32_t *)&buf[offset]);
o = *((uint32_t *)&buf[offset+4]);
if (ctx->machine_needs_byte_swap) {
v = byteswap4(v);
o = byteswap4(o);
}
if (v > 0 && o > 0) {
off_t cur_pos = readstat_lseek(fd, 0, SEEK_CUR);
if (cur_pos == -1) {
retval = READSTAT_ERROR_SEEK;
goto cleanup;
}
retval = dta_read_long_string(fd, ctx, v, o, &long_string);
if (retval != READSTAT_OK) {
goto cleanup;
}
value.v.string_value = long_string;
if (readstat_lseek(fd, cur_pos, SEEK_SET) == -1) {
retval = READSTAT_ERROR_SEEK;
goto cleanup;
}
}
} else if (value.type == READSTAT_TYPE_CHAR) {
char byte = buf[offset];
if (ctx->machine_is_twos_complement) {
byte = ones_to_twos_complement1(byte);
}
if (byte > DTA_MAX_CHAR) {
value.is_system_missing = 1;
if (byte > DTA_MISSING_CHAR) {
value.tag = 'a' + (byte - DTA_MISSING_CHAR_A);
}
}
value.v.char_value = byte;
} else if (value.type == READSTAT_TYPE_INT16) {
int16_t num = *((int16_t *)&buf[offset]);
if (ctx->machine_needs_byte_swap) {
num = byteswap2(num);
}
if (ctx->machine_is_twos_complement) {
num = ones_to_twos_complement2(num);
}
if (num > DTA_MAX_INT16) {
value.is_system_missing = 1;
if (num > DTA_MISSING_INT16) {
value.tag = 'a' + (num - DTA_MISSING_INT16_A);
}
}
value.v.i16_value = num;
} else if (value.type == READSTAT_TYPE_INT32) {
int32_t num = *((int32_t *)&buf[offset]);
if (ctx->machine_needs_byte_swap) {
num = byteswap4(num);
}
if (ctx->machine_is_twos_complement) {
num = ones_to_twos_complement4(num);
}
if (num > DTA_MAX_INT32) {
value.is_system_missing = 1;
if (num > DTA_MISSING_INT32) {
value.tag = 'a' + (num - DTA_MISSING_INT32_A);
}
}
value.v.i32_value = num;
} else if (value.type == READSTAT_TYPE_FLOAT) {
uint32_t num = *((uint32_t *)&buf[offset]);
float f_num = NAN;
if (ctx->machine_needs_byte_swap) {
num = byteswap4(num);
}
if (num > DTA_MAX_FLOAT) {
value.is_system_missing = 1;
if (num > DTA_MISSING_FLOAT) {
value.tag = 'a' + ((num - DTA_MISSING_FLOAT_A) >> 11);
}
} else {
static readstat_error_t dta_handle_rows(dta_ctx_t *ctx) {
readstat_io_t *io = ctx->io;
char *buf = NULL;
char str_buf[2048];
int i;
readstat_error_t retval = READSTAT_OK;
if ((buf = malloc(ctx->record_len)) == NULL) {
retval = READSTAT_ERROR_MALLOC;
goto cleanup;
}
for (i=0; i<ctx->row_limit; i++) {
if (io->read(buf, ctx->record_len, io->io_ctx) != ctx->record_len) {
retval = READSTAT_ERROR_READ;
goto cleanup;
}
int j;
off_t offset = 0;
for (j=0; j<ctx->nvar; j++) {
size_t max_len;
readstat_value_t value;
memset(&value, 0, sizeof(readstat_value_t));
value.type = dta_type_info(ctx->typlist[j], &max_len, ctx);
if (value.type == READSTAT_TYPE_STRING) {
readstat_convert(str_buf, sizeof(str_buf), &buf[offset], max_len, ctx->converter);
value.v.string_value = str_buf;
} else if (value.type == READSTAT_TYPE_STRING_REF) {
dta_strl_t key;
dta_interpret_strl_vo_bytes(ctx, (unsigned char *)&buf[offset], &key);
dta_strl_t **found = bsearch(&key, ctx->strls, ctx->strls_count, sizeof(dta_strl_t *), &dta_compare_strls);
if (found) {
value.v.string_value = (*found)->data;
}
value.type = READSTAT_TYPE_STRING;
} else if (value.type == READSTAT_TYPE_INT8) {
int8_t byte = buf[offset];
if (ctx->machine_is_twos_complement) {
byte = ones_to_twos_complement1(byte);
}
if (byte > ctx->max_int8) {
if (ctx->supports_tagged_missing && byte > DTA_113_MISSING_INT8) {
value.tag = 'a' + (byte - DTA_113_MISSING_INT8_A);
value.is_tagged_missing = 1;
} else {
value.is_system_missing = 1;
}
}
value.v.i8_value = byte;
} else if (value.type == READSTAT_TYPE_INT16) {
int16_t num = *((int16_t *)&buf[offset]);
if (ctx->machine_needs_byte_swap) {
num = byteswap2(num);
}
if (ctx->machine_is_twos_complement) {
num = ones_to_twos_complement2(num);
}
if (num > ctx->max_int16) {
if (ctx->supports_tagged_missing && num > DTA_113_MISSING_INT16) {
value.tag = 'a' + (num - DTA_113_MISSING_INT16_A);
value.is_tagged_missing = 1;
} else {
value.is_system_missing = 1;
}
}
value.v.i16_value = num;
} else if (value.type == READSTAT_TYPE_INT32) {
int32_t num = *((int32_t *)&buf[offset]);
if (ctx->machine_needs_byte_swap) {
num = byteswap4(num);
}
if (ctx->machine_is_twos_complement) {
num = ones_to_twos_complement4(num);
}
if (num > ctx->max_int32) {
if (ctx->supports_tagged_missing && num > DTA_113_MISSING_INT32) {
value.tag = 'a' + (num - DTA_113_MISSING_INT32_A);
value.is_tagged_missing = 1;
} else {
value.is_system_missing = 1;
}
}
value.v.i32_value = num;
} else if (value.type == READSTAT_TYPE_FLOAT) {
int32_t num = *((int32_t *)&buf[offset]);
float f_num = NAN;
if (ctx->machine_needs_byte_swap) {
num = byteswap4(num);
}
if (num > ctx->max_float) {
if (ctx->supports_tagged_missing && num > DTA_113_MISSING_FLOAT) {
value.tag = 'a' + ((num - DTA_113_MISSING_FLOAT_A) >> 11);
value.is_tagged_missing = 1;
} else {
value.is_system_missing = 1;
}
} else {
