static readstat_error_t dta_read_map(int fd, dta_ctx_t *ctx) {
if (!ctx->file_is_xmlish)
return 0;
readstat_error_t retval = READSTAT_OK;
if ((retval = dta_read_tag(fd, ctx, "<map>")) != READSTAT_OK) {
goto cleanup;
}
uint64_t map_buffer[14];
if (read(fd, map_buffer, sizeof(map_buffer)) != sizeof(map_buffer)) {
retval = READSTAT_ERROR_READ;
goto cleanup;
}
ctx->data_offset = ctx->machine_needs_byte_swap ? byteswap8(map_buffer[9]) : map_buffer[9];
ctx->strls_offset = ctx->machine_needs_byte_swap ? byteswap8(map_buffer[10]) : map_buffer[10];
ctx->value_labels_offset = ctx->machine_needs_byte_swap ? byteswap8(map_buffer[11]) : map_buffer[11];
if ((retval = dta_read_tag(fd, ctx, "</map>")) != READSTAT_OK) {
goto cleanup;
}
cleanup:
return retval;
}
static readstat_error_t dta_read_chunk(
dta_ctx_t *ctx,
const char *start_tag,
void *dst, size_t dst_len,
const char *end_tag) {
char *dst_buffer = (char *)dst;
readstat_io_t *io = ctx->io;
readstat_error_t retval = READSTAT_OK;
if ((retval = dta_read_tag(ctx, start_tag)) != READSTAT_OK)
goto cleanup;
if (io->read(dst_buffer, dst_len, io->io_ctx) != dst_len) {
retval = READSTAT_ERROR_READ;
goto cleanup;
}
if ((retval = dta_read_tag(ctx, end_tag)) != READSTAT_OK)
goto cleanup;
cleanup:
return retval;
}
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 {
readstat_error_t dta_read_xmlish_preamble(int fd, dta_ctx_t *ctx, dta_header_t *header) {
readstat_error_t retval = READSTAT_OK;
if ((retval = dta_read_tag(fd, ctx, "<stata_dta>")) != READSTAT_OK) {
goto cleanup;
}
if ((retval = dta_read_tag(fd, ctx, "<header>")) != READSTAT_OK) {
goto cleanup;
}
char ds_format[3];
if ((retval = dta_read_tag(fd, ctx, "<release>")) != READSTAT_OK) {
goto cleanup;
}
if (read(fd, ds_format, sizeof(ds_format)) != sizeof(ds_format)) {
retval = READSTAT_ERROR_READ;
goto cleanup;
}
header->ds_format = 100 * (ds_format[0] - '0') + 10 * (ds_format[1] - '0') + (ds_format[2] - '0');
if ((retval = dta_read_tag(fd, ctx, "</release>")) != READSTAT_OK) {
goto cleanup;
}
char byteorder[3];
if ((retval = dta_read_tag(fd, ctx, "<byteorder>")) != READSTAT_OK) {
goto cleanup;
}
if (read(fd, byteorder, sizeof(byteorder)) != sizeof(byteorder)) {
retval = READSTAT_ERROR_READ;
goto cleanup;
}
if (strncmp(byteorder, "MSF", 3) == 0) {
header->byteorder = DTA_HILO;
} else if (strncmp(byteorder, "LSF", 3) == 0) {
header->byteorder = DTA_LOHI;
} else {
retval = READSTAT_ERROR_PARSE;
goto cleanup;
}
if ((retval = dta_read_tag(fd, ctx, "</byteorder>")) != READSTAT_OK) {
goto cleanup;
}
if ((retval = dta_read_tag(fd, ctx, "<K>")) != READSTAT_OK) {
goto cleanup;
}
if (read(fd, &header->nvar, sizeof(int16_t)) != sizeof(int16_t)) {
retval = READSTAT_ERROR_READ;
goto cleanup;
}
if ((retval = dta_read_tag(fd, ctx, "</K>")) != READSTAT_OK) {
goto cleanup;
}
if ((retval = dta_read_tag(fd, ctx, "<N>")) != READSTAT_OK) {
goto cleanup;
}
if (read(fd, &header->nobs, sizeof(int32_t)) != sizeof(int32_t)) {
retval = READSTAT_ERROR_READ;
goto cleanup;
}
if (header->ds_format >= 118) {
/* Only support files < 4 billion rows for now */
if (header->byteorder == DTA_HILO) {
if (read(fd, &header->nobs, sizeof(int32_t)) != sizeof(int32_t)) {
retval = READSTAT_ERROR_READ;
goto cleanup;
}
} else {
if (readstat_lseek(fd, 4, SEEK_CUR) == -1) {
retval = READSTAT_ERROR_SEEK;
goto cleanup;
}
}
}
if ((retval = dta_read_tag(fd, ctx, "</N>")) != READSTAT_OK) {
goto cleanup;
}
cleanup:
return retval;
}
static readstat_error_t dta_read_long_string(int fd, dta_ctx_t *ctx, int v, int o, char **long_string_out) {
readstat_error_t retval = READSTAT_OK;
if (readstat_lseek(fd, ctx->strls_offset, SEEK_SET) != ctx->strls_offset) {
retval = READSTAT_ERROR_SEEK;
goto cleanup;
}
retval = dta_read_tag(fd, ctx, "<strls>");
if (retval != READSTAT_OK)
goto cleanup;
dta_gso_header_t header;
while (1) {
if (read(fd, &header, sizeof(dta_gso_header_t)) != sizeof(dta_gso_header_t)) {
retval = READSTAT_ERROR_READ;
goto cleanup;
}
if (strncmp(header.gso, "GSO", sizeof("GSO")-1) != 0) {
retval = READSTAT_ERROR_PARSE;
goto cleanup;
}
if (header.len <= 0) {
retval = READSTAT_ERROR_PARSE;
goto cleanup;
}
if (header.v == v && header.o == o) {
if (header.t == DTA_GSO_TYPE_BINARY) {
*long_string_out = NULL;
} else if (header.t == DTA_GSO_TYPE_ASCII) {
char *string_buf = malloc(header.len);
if (read(fd, string_buf, header.len) != header.len) {
free(string_buf);
retval = READSTAT_ERROR_READ;
goto cleanup;
}
if (string_buf[header.len-1] != '\0') {
free(string_buf);
retval = READSTAT_ERROR_PARSE;
goto cleanup;
}
*long_string_out = string_buf;
} else {
retval = READSTAT_ERROR_PARSE;
goto cleanup;
}
break;
} else {
if (readstat_lseek(fd, header.len, SEEK_CUR) == -1) {
retval = READSTAT_ERROR_SEEK;
goto cleanup;
}
}
}
cleanup:
return retval;
}
static readstat_error_t dta_read_descriptors(int fd, dta_ctx_t *ctx) {
if (dta_read_tag(fd, ctx, "<variable_types>") != READSTAT_OK)
return -1;
size_t buffer_len = ctx->nvar * ctx->typlist_entry_len;
unsigned char *buffer = malloc(buffer_len);
if (read(fd, buffer, buffer_len) != buffer_len) {
free(buffer);
return -1;
}
int i;
if (ctx->typlist_entry_len == 1) {
for (i=0; i<ctx->nvar; i++) {
ctx->typlist[i] = buffer[i];
}
} else if (ctx->typlist_entry_len == 2) {
memcpy(ctx->typlist, buffer, buffer_len);
if (ctx->machine_needs_byte_swap) {
for (i=0; i<ctx->nvar; i++) {
ctx->typlist[i] = byteswap2(ctx->typlist[i]);
}
}
}
free(buffer);
if (dta_read_tag(fd, ctx, "</variable_types>") != READSTAT_OK)
return -1;
if (dta_read_tag(fd, ctx, "<varnames>") != READSTAT_OK ||
read(fd, ctx->varlist, ctx->varlist_len) != ctx->varlist_len ||
dta_read_tag(fd, ctx, "</varnames>") != READSTAT_OK)
return -1;
if (dta_read_tag(fd, ctx, "<sortlist>") != READSTAT_OK ||
read(fd, ctx->srtlist, ctx->srtlist_len) != ctx->srtlist_len ||
dta_read_tag(fd, ctx, "</sortlist>") != READSTAT_OK)
return -1;
if (dta_read_tag(fd, ctx, "<formats>") != READSTAT_OK ||
read(fd, ctx->fmtlist, ctx->fmtlist_len) != ctx->fmtlist_len ||
dta_read_tag(fd, ctx, "</formats>") != READSTAT_OK)
return -1;
if (dta_read_tag(fd, ctx, "<value_label_names>") != READSTAT_OK ||
read(fd, ctx->lbllist, ctx->lbllist_len) != ctx->lbllist_len ||
dta_read_tag(fd, ctx, "</value_label_names>") != READSTAT_OK)
return -1;
if (dta_read_tag(fd, ctx, "<variable_labels>") != READSTAT_OK ||
read(fd, ctx->variable_labels, ctx->variable_labels_len) != ctx->variable_labels_len ||
dta_read_tag(fd, ctx, "</variable_labels>") != READSTAT_OK)
return -1;
return 0;
}
static readstat_error_t dta_read_strls(dta_ctx_t *ctx) {
if (!ctx->file_is_xmlish)
return READSTAT_OK;
readstat_error_t retval = READSTAT_OK;
readstat_io_t *io = ctx->io;
if (io->seek(ctx->strls_offset, READSTAT_SEEK_SET, io->io_ctx) == -1) {
retval = READSTAT_ERROR_SEEK;
goto cleanup;
}
retval = dta_read_tag(ctx, "<strls>");
if (retval != READSTAT_OK)
goto cleanup;
ctx->strls_capacity = 100;
ctx->strls = malloc(ctx->strls_capacity * sizeof(dta_strl_t *));
while (1) {
char tag[3];
if (io->read(tag, sizeof(tag), io->io_ctx) != sizeof(tag)) {
retval = READSTAT_ERROR_READ;
goto cleanup;
}
if (memcmp(tag, "GSO", sizeof(tag)) == 0) {
dta_strl_t strl;
retval = dta_read_strl(ctx, &strl);
if (retval != READSTAT_OK)
goto cleanup;
if (strl.type != DTA_GSO_TYPE_ASCII)
continue;
if (ctx->strls_count == ctx->strls_capacity) {
ctx->strls_capacity *= 2;
ctx->strls = realloc(ctx->strls, sizeof(dta_strl_t *) * ctx->strls_capacity);
}
dta_strl_t *strl_ptr = malloc(sizeof(dta_strl_t) + strl.len);
memcpy(strl_ptr, &strl, sizeof(dta_strl_t));
ctx->strls[ctx->strls_count++] = strl_ptr;
if (io->read(&strl_ptr->data[0], strl_ptr->len, io->io_ctx) != strl_ptr->len) {
retval = READSTAT_ERROR_READ;
goto cleanup;
}
} else if (memcmp(tag, "</s", sizeof(tag)) == 0) {
retval = dta_read_tag(ctx, "trls>");
if (retval != READSTAT_OK)
goto cleanup;
break;
} else {
retval = READSTAT_ERROR_PARSE;
goto cleanup;
}
}
cleanup:
return retval;
}
static readstat_error_t dta_read_expansion_fields(dta_ctx_t *ctx) {
readstat_error_t retval = READSTAT_OK;
readstat_io_t *io = ctx->io;
char *buffer = NULL;
if (ctx->expansion_len_len == 0)
return READSTAT_OK;
if (ctx->file_is_xmlish && !ctx->note_handler) {
if (io->seek(ctx->data_offset, READSTAT_SEEK_SET, io->io_ctx) == -1) {
return READSTAT_ERROR_SEEK;
}
return READSTAT_OK;
}
retval = dta_read_tag(ctx, "<characteristics>");
if (retval != READSTAT_OK)
goto cleanup;
while (1) {
size_t len;
char data_type;
if (ctx->file_is_xmlish) {
char start[4];
if (io->read(start, sizeof(start), io->io_ctx) != sizeof(start)) {
retval = READSTAT_ERROR_READ;
goto cleanup;
}
if (memcmp(start, "</ch", sizeof(start)) == 0) {
retval = dta_read_tag(ctx, "aracteristics>");
if (retval != READSTAT_OK)
goto cleanup;
break;
} else if (memcmp(start, "<ch>", sizeof(start)) != 0) {
retval = READSTAT_ERROR_PARSE;
goto cleanup;
}
data_type = 1;
} else {
if (io->read(&data_type, 1, io->io_ctx) != 1) {
retval = READSTAT_ERROR_READ;
goto cleanup;
}
}
if (ctx->expansion_len_len == 2) {
int16_t len16;
if (io->read(&len16, sizeof(int16_t), io->io_ctx) != sizeof(int16_t)) {
retval = READSTAT_ERROR_READ;
goto cleanup;
}
len = ctx->machine_needs_byte_swap ? byteswap2(len16) : len16;
} else {
int32_t len32;
if (io->read(&len32, sizeof(int32_t), io->io_ctx) != sizeof(int32_t)) {
retval = READSTAT_ERROR_READ;
goto cleanup;
}
len = ctx->machine_needs_byte_swap ? byteswap2(len32) : len32;
}
if (data_type == 0 && len == 0)
break;
if (data_type != 1 || len > (1<<20)) {
retval = READSTAT_ERROR_NOTE_IS_TOO_LONG;
goto cleanup;
}
if (ctx->note_handler && len >= 2 * ctx->ch_metadata_len) {
buffer = realloc(buffer, len + 1);
buffer[len] = '\0';
if (io->read(buffer, len, io->io_ctx) != len) {
retval = READSTAT_ERROR_READ;
goto cleanup;
}
int index = 0;
if (strncmp(&buffer[0], "_dta", 4) == 0 &&
sscanf(&buffer[ctx->ch_metadata_len], "note%d", &index) == 1) {
if (ctx->note_handler(index, &buffer[2*ctx->ch_metadata_len], ctx->user_ctx)) {
retval = READSTAT_ERROR_USER_ABORT;
goto cleanup;
}
}
} else {
if (io->seek(len, READSTAT_SEEK_CUR, io->io_ctx) == -1) {
retval = READSTAT_ERROR_SEEK;
goto cleanup;
}
}
retval = dta_read_tag(ctx, "</ch>");
if (retval != READSTAT_OK)
goto cleanup;
}
cleanup:
if (buffer)
free(buffer);
return retval;
}
