2 * This file is part of the libsigrok project.
4 * Copyright (C) 2013 Marc Schink <sigrok-dev@marcschink.de>
6 * This program is free software: you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License as published by
8 * the Free Software Foundation, either version 3 of the License, or
9 * (at your option) any later version.
11 * This program is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 * GNU General Public License for more details.
16 * You should have received a copy of the GNU General Public License
17 * along with this program. If not, see <http://www.gnu.org/licenses/>.
26 #include <libsigrok/libsigrok.h>
27 #include "libsigrok-internal.h"
29 #define LOG_PREFIX "input/csv"
31 #define CHUNK_SIZE (4 * 1024 * 1024)
34 * The CSV input module has the following options:
36 * single-column: Specifies the column number which stores the sample data for
37 * single column mode and enables single column mode. Multi
38 * column mode is used if this parameter is omitted.
40 * numchannels: Specifies the number of channels to use. In multi column mode
41 * the number of channels are the number of columns and in single
42 * column mode the number of bits (LSB first) beginning at
45 * delimiter: Specifies the delimiter for columns. Must be at least one
46 * character. Comma is used as default delimiter.
48 * format: Specifies the format of the sample data in single column mode.
49 * Available formats are: 'bin', 'hex' and 'oct'. The binary
50 * format is used by default. This option has no effect in multi
53 * comment: Specifies the prefix character(s) for comments. No prefix
54 * characters are used by default which disables removing of
57 * samplerate: Samplerate which the sample data was captured with. Default
60 * first-channel: Column number of the first channel in multi column mode and
61 * position of the bit for the first channel in single column mode.
64 * header: Determines if the first line should be treated as header
65 * and used for channel names in multi column mode. Empty header
66 * names will be replaced by the channel number. If enabled in
67 * single column mode the first line will be skipped. Usage of
68 * header is disabled by default.
70 * startline: Line number to start processing sample data. Must be greater
71 * than 0. The default line number to start processing is 1.
77 * - Determine how the text line handling can get improved, regarding
78 * all of robustness and flexibility and correctness.
79 * - The current implementation splits on "any run of CR and LF". Which
80 * translates to: Line numbers are wrong in the presence of empty
81 * lines in the input stream. See below for an (expensive) fix.
82 * - Dropping support for CR style end-of-line markers could improve
83 * the situation a lot. Code could search for and split on LF, and
84 * trim optional trailing CR. This would result in proper support
85 * for CRLF (Windows) as well as LF (Unix), and allow for correct
87 * - When support for CR-only line termination cannot get dropped,
88 * then the current implementation is inappropriate. Currently the
89 * input stream is scanned for the first occurance of either of the
90 * supported termination styles (which is good). For the remaining
91 * session a consistent encoding of the text lines is assumed (which
93 * - When line numbers need to be correct and reliable, _and_ the full
94 * set of previously supported line termination sequences are required,
95 * and potentially more are to get added for improved compatibility
96 * with more platforms or generators, then the current approach of
97 * splitting on runs of termination characters needs to get replaced,
98 * by the more expensive approach to scan for and count the initially
99 * determined termination sequence.
101 * - Add support for analog input data? (optional)
102 * - Needs a syntax first for user specs which channels (columns) are
103 * logic and which are analog. May need heuristics(?) to guess from
104 * input data in the absence of user provided specs.
107 /* Single column formats. */
117 /* Current selected samplerate. */
120 /* Number of channels. */
121 unsigned int num_channels;
123 /* Column delimiter character(s). */
126 /* Comment prefix character(s). */
129 /* Termination character(s) used in current stream. */
132 /* Determines if sample data is stored in multiple columns. */
133 gboolean multi_column_mode;
135 /* Column number of the sample data in single column mode. */
136 unsigned int single_column;
139 * Number of the first column to parse. Equivalent to the number of the
140 * first channel in multi column mode and the single column number in
141 * single column mode.
143 unsigned int first_column;
146 * Column number of the first channel in multi column mode and position of
147 * the bit for the first channel in single column mode.
149 unsigned int first_channel;
151 /* Line number to start processing. */
155 * Determines if the first line should be treated as header and used for
156 * channel names in multi column mode.
160 /* Format sample data is stored in single column mode. */
163 size_t sample_unit_size; /**!< Byte count for a single sample. */
164 uint8_t *sample_buffer; /**!< Buffer for a single sample. */
166 uint8_t *datafeed_buffer; /**!< Queue for datafeed submission. */
167 size_t datafeed_buf_size;
168 size_t datafeed_buf_fill;
170 /* Current line number. */
173 /* List of previously created sigrok channels. */
174 GSList *prev_sr_channels;
177 static void strip_comment(char *buf, const GString *prefix)
184 if ((ptr = strstr(buf, prefix->str)))
188 static int parse_binstr(const char *str, struct context *inc)
192 length = strlen(str);
195 sr_err("Column %u in line %zu is empty.", inc->single_column,
200 /* Clear buffer in order to set bits only. */
201 memset(inc->sample_buffer, 0, inc->sample_unit_size);
203 i = inc->first_channel;
205 for (j = 0; i < length && j < inc->num_channels; i++, j++) {
206 if (str[length - i - 1] == '1') {
207 inc->sample_buffer[j / 8] |= (1 << (j % 8));
208 } else if (str[length - i - 1] != '0') {
209 sr_err("Invalid value '%s' in column %u in line %zu.",
210 str, inc->single_column, inc->line_number);
218 static int parse_hexstr(const char *str, struct context *inc)
220 gsize i, j, k, length;
224 length = strlen(str);
227 sr_err("Column %u in line %zu is empty.", inc->single_column,
232 /* Clear buffer in order to set bits only. */
233 memset(inc->sample_buffer, 0, inc->sample_unit_size);
235 /* Calculate the position of the first hexadecimal digit. */
236 i = inc->first_channel / 4;
238 for (j = 0; i < length && j < inc->num_channels; i++) {
239 c = str[length - i - 1];
241 if (!g_ascii_isxdigit(c)) {
242 sr_err("Invalid value '%s' in column %u in line %zu.",
243 str, inc->single_column, inc->line_number);
247 value = g_ascii_xdigit_value(c);
249 k = (inc->first_channel + j) % 4;
251 for (; j < inc->num_channels && k < 4; k++) {
252 if (value & (1 << k))
253 inc->sample_buffer[j / 8] |= (1 << (j % 8));
262 static int parse_octstr(const char *str, struct context *inc)
264 gsize i, j, k, length;
268 length = strlen(str);
271 sr_err("Column %u in line %zu is empty.", inc->single_column,
276 /* Clear buffer in order to set bits only. */
277 memset(inc->sample_buffer, 0, inc->sample_unit_size);
279 /* Calculate the position of the first octal digit. */
280 i = inc->first_channel / 3;
282 for (j = 0; i < length && j < inc->num_channels; i++) {
283 c = str[length - i - 1];
285 if (c < '0' || c > '7') {
286 sr_err("Invalid value '%s' in column %u in line %zu.",
287 str, inc->single_column, inc->line_number);
291 value = g_ascii_xdigit_value(c);
293 k = (inc->first_channel + j) % 3;
295 for (; j < inc->num_channels && k < 3; k++) {
296 if (value & (1 << k))
297 inc->sample_buffer[j / 8] |= (1 << (j % 8));
306 static char **parse_line(char *buf, struct context *inc, int max_columns)
308 const char *str, *remainder;
319 str = strstr(remainder, inc->delimiter->str);
321 while (str && max_columns) {
322 if (n >= inc->first_column) {
323 column = g_strndup(remainder, str - remainder);
324 list = g_slist_prepend(list, g_strstrip(column));
330 remainder = str + inc->delimiter->len;
331 str = strstr(remainder, inc->delimiter->str);
335 if (buf[0] && max_columns && n >= inc->first_column) {
336 column = g_strdup(remainder);
337 list = g_slist_prepend(list, g_strstrip(column));
341 if (!(columns = g_try_new(char *, k + 1)))
346 for (l = list; l; l = l->next)
347 columns[k--] = l->data;
354 static int parse_multi_columns(char **columns, struct context *inc)
359 /* Clear buffer in order to set bits only. */
360 memset(inc->sample_buffer, 0, inc->sample_unit_size);
362 for (i = 0; i < inc->num_channels; i++) {
364 if (column[0] == '1') {
365 inc->sample_buffer[i / 8] |= (1 << (i % 8));
366 } else if (!strlen(column)) {
367 sr_err("Column %zu in line %zu is empty.",
368 inc->first_channel + i, inc->line_number);
370 } else if (column[0] != '0') {
371 sr_err("Invalid value '%s' in column %zu in line %zu.",
372 column, inc->first_channel + i,
381 static int parse_single_column(const char *column, struct context *inc)
387 switch (inc->format) {
389 res = parse_binstr(column, inc);
392 res = parse_hexstr(column, inc);
395 res = parse_octstr(column, inc);
402 static int flush_samples(const struct sr_input *in)
405 struct sr_datafeed_packet packet;
406 struct sr_datafeed_logic logic;
410 if (!inc->datafeed_buf_fill)
413 memset(&packet, 0, sizeof(packet));
414 memset(&logic, 0, sizeof(logic));
415 packet.type = SR_DF_LOGIC;
416 packet.payload = &logic;
417 logic.unitsize = inc->sample_unit_size;
418 logic.length = inc->datafeed_buf_fill;
419 logic.data = inc->datafeed_buffer;
421 rc = sr_session_send(in->sdi, &packet);
425 inc->datafeed_buf_fill = 0;
429 static int queue_samples(const struct sr_input *in)
436 inc->datafeed_buf_fill += inc->sample_unit_size;
437 if (inc->datafeed_buf_fill == inc->datafeed_buf_size) {
438 rc = flush_samples(in);
442 inc->sample_buffer = &inc->datafeed_buffer[inc->datafeed_buf_fill];
446 static int init(struct sr_input *in, GHashTable *options)
451 in->sdi = g_malloc0(sizeof(struct sr_dev_inst));
452 in->priv = inc = g_malloc0(sizeof(struct context));
454 inc->single_column = g_variant_get_int32(g_hash_table_lookup(options, "single-column"));
455 inc->multi_column_mode = inc->single_column == 0;
457 inc->num_channels = g_variant_get_int32(g_hash_table_lookup(options, "numchannels"));
459 inc->delimiter = g_string_new(g_variant_get_string(
460 g_hash_table_lookup(options, "delimiter"), NULL));
461 if (inc->delimiter->len == 0) {
462 sr_err("Delimiter must be at least one character.");
466 s = g_variant_get_string(g_hash_table_lookup(options, "format"), NULL);
467 if (!g_ascii_strncasecmp(s, "bin", 3)) {
468 inc->format = FORMAT_BIN;
469 } else if (!g_ascii_strncasecmp(s, "hex", 3)) {
470 inc->format = FORMAT_HEX;
471 } else if (!g_ascii_strncasecmp(s, "oct", 3)) {
472 inc->format = FORMAT_OCT;
474 sr_err("Invalid format: '%s'", s);
478 inc->comment = g_string_new(g_variant_get_string(
479 g_hash_table_lookup(options, "comment"), NULL));
480 if (g_string_equal(inc->comment, inc->delimiter)) {
481 /* That's never going to work. Likely the result of the user
482 * setting the delimiter to ; -- the default comment. Clearing
483 * the comment setting will work in that case. */
484 g_string_truncate(inc->comment, 0);
487 inc->samplerate = g_variant_get_uint64(g_hash_table_lookup(options, "samplerate"));
489 inc->first_channel = g_variant_get_int32(g_hash_table_lookup(options, "first-channel"));
491 inc->header = g_variant_get_boolean(g_hash_table_lookup(options, "header"));
493 inc->start_line = g_variant_get_int32(g_hash_table_lookup(options, "startline"));
494 if (inc->start_line < 1) {
495 sr_err("Invalid start line %zu.", inc->start_line);
499 if (inc->multi_column_mode)
500 inc->first_column = inc->first_channel;
502 inc->first_column = inc->single_column;
504 if (!inc->multi_column_mode && !inc->num_channels) {
505 sr_err("Number of channels needs to be specified in single column mode.");
513 * Check the channel list for consistency across file re-import. See
514 * the VCD input module for more details and motivation.
517 static void keep_header_for_reread(const struct sr_input *in)
522 g_slist_free_full(inc->prev_sr_channels, sr_channel_free_cb);
523 inc->prev_sr_channels = in->sdi->channels;
524 in->sdi->channels = NULL;
527 static int check_header_in_reread(const struct sr_input *in)
536 if (!inc->prev_sr_channels)
539 if (sr_channel_lists_differ(inc->prev_sr_channels, in->sdi->channels)) {
540 sr_err("Channel list change not supported for file re-read.");
543 g_slist_free_full(in->sdi->channels, sr_channel_free_cb);
544 in->sdi->channels = inc->prev_sr_channels;
545 inc->prev_sr_channels = NULL;
550 static const char *delim_set = "\r\n";
552 static const char *get_line_termination(GString *buf)
557 if (g_strstr_len(buf->str, buf->len, "\r\n"))
559 else if (memchr(buf->str, '\n', buf->len))
561 else if (memchr(buf->str, '\r', buf->len))
567 static int initial_parse(const struct sr_input *in, GString *buf)
570 GString *channel_name;
571 unsigned int num_columns, i;
572 size_t line_number, l;
574 char **lines, *line, **columns, *column;
581 lines = g_strsplit_set(buf->str, delim_set, 0);
582 for (l = 0; lines[l]; l++) {
585 if (inc->start_line > line_number) {
586 sr_spew("Line %zu skipped.", line_number);
589 if (line[0] == '\0') {
590 sr_spew("Blank line %zu skipped.", line_number);
593 strip_comment(line, inc->comment);
594 if (line[0] == '\0') {
595 sr_spew("Comment-only line %zu skipped.", line_number);
599 /* Reached first proper line. */
603 /* Not enough data for a proper line yet. */
609 * In order to determine the number of columns parse the current line
610 * without limiting the number of columns.
612 columns = parse_line(line, inc, -1);
614 sr_err("Error while parsing line %zu.", line_number);
618 num_columns = g_strv_length(columns);
620 /* Ensure that the first column is not out of bounds. */
622 sr_err("Column %u in line %zu is out of bounds.",
623 inc->first_column, line_number);
628 if (inc->multi_column_mode) {
630 * Detect the number of channels in multi column mode
631 * automatically if not specified.
633 if (!inc->num_channels) {
634 inc->num_channels = num_columns;
635 sr_dbg("Number of auto-detected channels: %u.",
640 * Ensure that the number of channels does not exceed the number
641 * of columns in multi column mode.
643 if (num_columns < inc->num_channels) {
644 sr_err("Not enough columns for desired number of channels in line %zu.",
651 channel_name = g_string_sized_new(64);
652 for (i = 0; i < inc->num_channels; i++) {
654 if (inc->header && inc->multi_column_mode && column[0] != '\0')
655 g_string_assign(channel_name, column);
657 g_string_printf(channel_name, "%u", i);
658 sr_channel_new(in->sdi, i, SR_CHANNEL_LOGIC, TRUE, channel_name->str);
660 g_string_free(channel_name, TRUE);
661 if (!check_header_in_reread(in)) {
667 * Calculate the minimum buffer size to store the set of samples
668 * of all channels (unit size). Determine a larger buffer size
669 * for datafeed submission that is a multiple of the unit size.
670 * Allocate the larger buffer, and have the "sample buffer" point
671 * to a location within that large buffer.
673 inc->sample_unit_size = (inc->num_channels + 7) / 8;
674 inc->datafeed_buf_size = CHUNK_SIZE;
675 inc->datafeed_buf_size *= inc->sample_unit_size;
676 inc->datafeed_buffer = g_malloc(inc->datafeed_buf_size);
677 inc->datafeed_buf_fill = 0;
678 inc->sample_buffer = &inc->datafeed_buffer[inc->datafeed_buf_fill];
689 * Gets called from initial_receive(), which runs until the end-of-line
690 * encoding of the input stream could get determined. Assumes that this
691 * routine receives enough buffered initial input data to either see the
692 * BOM when there is one, or that no BOM will follow when a text line
693 * termination sequence was seen. Silently drops the UTF-8 BOM sequence
694 * from the input buffer if one was seen. Does not care to protect
695 * against multiple execution or dropping the BOM multiple times --
696 * there should be at most one in the input stream.
698 static void initial_bom_check(const struct sr_input *in)
700 static const char *utf8_bom = "\xef\xbb\xbf";
702 if (in->buf->len < strlen(utf8_bom))
704 if (strncmp(in->buf->str, utf8_bom, strlen(utf8_bom)) != 0)
706 g_string_erase(in->buf, 0, strlen(utf8_bom));
709 static int initial_receive(const struct sr_input *in)
715 const char *termination;
717 initial_bom_check(in);
721 termination = get_line_termination(in->buf);
723 /* Don't have a full line yet. */
726 p = g_strrstr_len(in->buf->str, in->buf->len, termination);
728 /* Don't have a full line yet. */
730 len = p - in->buf->str - 1;
731 new_buf = g_string_new_len(in->buf->str, len);
732 g_string_append_c(new_buf, '\0');
734 inc->termination = g_strdup(termination);
736 if (in->buf->str[0] != '\0')
737 ret = initial_parse(in, new_buf);
741 g_string_free(new_buf, TRUE);
746 static int process_buffer(struct sr_input *in, gboolean is_eof)
748 struct sr_datafeed_packet packet;
749 struct sr_datafeed_meta meta;
750 struct sr_config *src;
754 int max_columns, ret, l;
755 char *p, **lines, *line, **columns;
759 std_session_send_df_header(in->sdi);
761 if (inc->samplerate) {
762 packet.type = SR_DF_META;
763 packet.payload = &meta;
764 samplerate = inc->samplerate;
765 src = sr_config_new(SR_CONF_SAMPLERATE, g_variant_new_uint64(samplerate));
766 meta.config = g_slist_append(NULL, src);
767 sr_session_send(in->sdi, &packet);
768 g_slist_free(meta.config);
775 /* Limit the number of columns to parse. */
776 if (inc->multi_column_mode)
777 max_columns = inc->num_channels;
782 * Consider empty input non-fatal. Keep accumulating input until
783 * at least one full text line has become available. Grab the
784 * maximum amount of accumulated data that consists of full text
785 * lines, and process what has been received so far, leaving not
786 * yet complete lines for the next invocation.
788 * Enforce that all previously buffered data gets processed in
789 * the "EOF" condition. Do not insist in the presence of the
790 * termination sequence for the last line (may often be missing
791 * on Windows). A present termination sequence will just result
792 * in the "execution of an empty line", and does not harm.
797 p = in->buf->str + in->buf->len;
799 p = g_strrstr_len(in->buf->str, in->buf->len, inc->termination);
803 p += strlen(inc->termination);
805 g_strstrip(in->buf->str);
808 lines = g_strsplit_set(in->buf->str, delim_set, 0);
809 for (l = 0; lines[l]; l++) {
812 if (line[0] == '\0') {
813 sr_spew("Blank line %zu skipped.", inc->line_number);
817 /* Remove trailing comment. */
818 strip_comment(line, inc->comment);
819 if (line[0] == '\0') {
820 sr_spew("Comment-only line %zu skipped.", inc->line_number);
824 /* Skip the header line, its content was used as the channel names. */
826 sr_spew("Header line %zu skipped.", inc->line_number);
831 columns = parse_line(line, inc, max_columns);
833 sr_err("Error while parsing line %zu.", inc->line_number);
837 num_columns = g_strv_length(columns);
839 sr_err("Column %u in line %zu is out of bounds.",
840 inc->first_column, inc->line_number);
846 * Ensure that the number of channels does not exceed the number
847 * of columns in multi column mode.
849 if (inc->multi_column_mode && num_columns < inc->num_channels) {
850 sr_err("Not enough columns for desired number of channels in line %zu.",
857 if (inc->multi_column_mode)
858 ret = parse_multi_columns(columns, inc);
860 ret = parse_single_column(columns[0], inc);
867 /* Send sample data to the session bus. */
868 ret = queue_samples(in);
870 sr_err("Sending samples failed.");
879 g_string_erase(in->buf, 0, p - in->buf->str);
884 static int receive(struct sr_input *in, GString *buf)
889 g_string_append_len(in->buf, buf->str, buf->len);
892 if (!inc->termination) {
893 ret = initial_receive(in);
894 if (ret == SR_ERR_NA)
895 /* Not enough data yet. */
897 else if (ret != SR_OK)
900 /* sdi is ready, notify frontend. */
901 in->sdi_ready = TRUE;
905 ret = process_buffer(in, FALSE);
910 static int end(struct sr_input *in)
916 ret = process_buffer(in, TRUE);
922 ret = flush_samples(in);
928 std_session_send_df_end(in->sdi);
933 static void cleanup(struct sr_input *in)
937 keep_header_for_reread(in);
941 g_free(inc->termination);
942 inc->termination = NULL;
943 g_free(inc->datafeed_buffer);
944 inc->datafeed_buffer = NULL;
947 static int reset(struct sr_input *in)
949 struct context *inc = in->priv;
952 inc->started = FALSE;
953 g_string_truncate(in->buf, 0);
958 static struct sr_option options[] = {
959 { "single-column", "Single column", "Enable single-column mode, using the specified column (>= 1); 0: multi-col. mode", NULL, NULL },
960 { "numchannels", "Number of logic channels", "The number of (logic) channels (single-col. mode: number of bits beginning at 'first channel', LSB-first)", NULL, NULL },
961 { "delimiter", "Column delimiter", "The column delimiter (>= 1 characters)", NULL, NULL },
962 { "format", "Data format (single-col. mode)", "The numeric format of the data (single-col. mode): bin, hex, oct", NULL, NULL },
963 { "comment", "Comment character(s)", "The comment prefix character(s)", NULL, NULL },
964 { "samplerate", "Samplerate (Hz)", "The sample rate (used during capture) in Hz", NULL, NULL },
965 { "first-channel", "First channel", "The column number of the first channel (multi-col. mode); bit position for the first channel (single-col. mode)", NULL, NULL },
966 { "header", "Interpret first line as header (multi-col. mode)", "Treat the first line as header with channel names (multi-col. mode)", NULL, NULL },
967 { "startline", "Start line", "The line number at which to start processing samples (>= 1)", NULL, NULL },
971 static const struct sr_option *get_options(void)
975 if (!options[0].def) {
976 options[0].def = g_variant_ref_sink(g_variant_new_int32(0));
977 options[1].def = g_variant_ref_sink(g_variant_new_int32(0));
978 options[2].def = g_variant_ref_sink(g_variant_new_string(","));
979 options[3].def = g_variant_ref_sink(g_variant_new_string("bin"));
981 l = g_slist_append(l, g_variant_ref_sink(g_variant_new_string("bin")));
982 l = g_slist_append(l, g_variant_ref_sink(g_variant_new_string("hex")));
983 l = g_slist_append(l, g_variant_ref_sink(g_variant_new_string("oct")));
984 options[3].values = l;
985 options[4].def = g_variant_ref_sink(g_variant_new_string(";"));
986 options[5].def = g_variant_ref_sink(g_variant_new_uint64(0));
987 options[6].def = g_variant_ref_sink(g_variant_new_int32(0));
988 options[7].def = g_variant_ref_sink(g_variant_new_boolean(FALSE));
989 options[8].def = g_variant_ref_sink(g_variant_new_int32(1));
995 SR_PRIV struct sr_input_module input_csv = {
998 .desc = "Comma-separated values",
999 .exts = (const char*[]){"csv", NULL},
1000 .options = get_options,