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/>.
24 #include <libsigrok/libsigrok.h>
25 #include "libsigrok-internal.h"
27 #define LOG_PREFIX "input/csv"
30 * The CSV input module has the following options:
32 * single-column: Specifies the column number which stores the sample data for
33 * single column mode and enables single column mode. Multi
34 * column mode is used if this parameter is omitted.
36 * numchannels: Specifies the number of channels to use. In multi column mode
37 * the number of channels are the number of columns and in single
38 * column mode the number of bits (LSB first) beginning at
41 * delimiter: Specifies the delimiter for columns. Must be at least one
42 * character. Comma is used as default delimiter.
44 * format: Specifies the format of the sample data in single column mode.
45 * Available formats are: 'bin', 'hex' and 'oct'. The binary
46 * format is used by default. This option has no effect in multi
49 * comment: Specifies the prefix character(s) for comments. No prefix
50 * characters are used by default which disables removing of
53 * samplerate: Samplerate which the sample data was captured with. Default
56 * first-channel: Column number of the first channel in multi column mode and
57 * position of the bit for the first channel in single column mode.
60 * header: Determines if the first line should be treated as header
61 * and used for channel names in multi column mode. Empty header
62 * names will be replaced by the channel number. If enabled in
63 * single column mode the first line will be skipped. Usage of
64 * header is disabled by default.
66 * startline: Line number to start processing sample data. Must be greater
67 * than 0. The default line number to start processing is 1.
73 * - Determine how the text line handling can get improved, regarding
74 * all of robustness and flexibility and correctness.
75 * - The current implementation splits on "any run of CR and LF". Which
76 * translates to: Line numbers are wrong in the presence of empty
77 * lines in the input stream. See below for an (expensive) fix.
78 * - Dropping support for CR style end-of-line markers could improve
79 * the situation a lot. Code could search for and split on LF, and
80 * trim optional trailing CR. This would result in proper support
81 * for CRLF (Windows) as well as LF (Unix), and allow for correct
83 * - When support for CR-only line termination cannot get dropped,
84 * then the current implementation is inappropriate. Currently the
85 * input stream is scanned for the first occurance of either of the
86 * supported termination styles (which is good). For the remaining
87 * session a consistent encoding of the text lines is assumed (which
89 * - When line numbers need to be correct and reliable, _and_ the full
90 * set of previously supported line termination sequences are required,
91 * and potentially more are to get added for improved compatibility
92 * with more platforms or generators, then the current approach of
93 * splitting on runs of termination characters needs to get replaced,
94 * by the more expensive approach to scan for and count the initially
95 * determined termination sequence.
97 * - Add support for analog input data? (optional)
98 * - Needs a syntax first for user specs which channels (columns) are
99 * logic and which are analog. May need heuristics(?) to guess from
100 * input data in the absence of user provided specs.
103 /* Single column formats. */
113 /* Current selected samplerate. */
116 /* Number of channels. */
117 unsigned int num_channels;
119 /* Column delimiter character(s). */
122 /* Comment prefix character(s). */
125 /* Termination character(s) used in current stream. */
128 /* Determines if sample data is stored in multiple columns. */
129 gboolean multi_column_mode;
131 /* Column number of the sample data in single column mode. */
132 unsigned int single_column;
135 * Number of the first column to parse. Equivalent to the number of the
136 * first channel in multi column mode and the single column number in
137 * single column mode.
139 unsigned int first_column;
142 * Column number of the first channel in multi column mode and position of
143 * the bit for the first channel in single column mode.
145 unsigned int first_channel;
147 /* Line number to start processing. */
151 * Determines if the first line should be treated as header and used for
152 * channel names in multi column mode.
156 /* Format sample data is stored in single column mode. */
159 /* Size of the sample buffer. */
160 size_t sample_buffer_size;
162 /* Buffer to store sample data. */
163 uint8_t *sample_buffer;
165 /* Current line number. */
169 static void strip_comment(char *buf, const GString *prefix)
176 if ((ptr = strstr(buf, prefix->str)))
180 static int parse_binstr(const char *str, struct context *inc)
184 length = strlen(str);
187 sr_err("Column %u in line %zu is empty.", inc->single_column,
192 /* Clear buffer in order to set bits only. */
193 memset(inc->sample_buffer, 0, (inc->num_channels + 7) >> 3);
195 i = inc->first_channel;
197 for (j = 0; i < length && j < inc->num_channels; i++, j++) {
198 if (str[length - i - 1] == '1') {
199 inc->sample_buffer[j / 8] |= (1 << (j % 8));
200 } else if (str[length - i - 1] != '0') {
201 sr_err("Invalid value '%s' in column %u in line %zu.",
202 str, inc->single_column, inc->line_number);
210 static int parse_hexstr(const char *str, struct context *inc)
212 gsize i, j, k, length;
216 length = strlen(str);
219 sr_err("Column %u in line %zu is empty.", inc->single_column,
224 /* Clear buffer in order to set bits only. */
225 memset(inc->sample_buffer, 0, (inc->num_channels + 7) >> 3);
227 /* Calculate the position of the first hexadecimal digit. */
228 i = inc->first_channel / 4;
230 for (j = 0; i < length && j < inc->num_channels; i++) {
231 c = str[length - i - 1];
233 if (!g_ascii_isxdigit(c)) {
234 sr_err("Invalid value '%s' in column %u in line %zu.",
235 str, inc->single_column, inc->line_number);
239 value = g_ascii_xdigit_value(c);
241 k = (inc->first_channel + j) % 4;
243 for (; j < inc->num_channels && k < 4; k++) {
244 if (value & (1 << k))
245 inc->sample_buffer[j / 8] |= (1 << (j % 8));
254 static int parse_octstr(const char *str, struct context *inc)
256 gsize i, j, k, length;
260 length = strlen(str);
263 sr_err("Column %u in line %zu is empty.", inc->single_column,
268 /* Clear buffer in order to set bits only. */
269 memset(inc->sample_buffer, 0, (inc->num_channels + 7) >> 3);
271 /* Calculate the position of the first octal digit. */
272 i = inc->first_channel / 3;
274 for (j = 0; i < length && j < inc->num_channels; i++) {
275 c = str[length - i - 1];
277 if (c < '0' || c > '7') {
278 sr_err("Invalid value '%s' in column %u in line %zu.",
279 str, inc->single_column, inc->line_number);
283 value = g_ascii_xdigit_value(c);
285 k = (inc->first_channel + j) % 3;
287 for (; j < inc->num_channels && k < 3; k++) {
288 if (value & (1 << k))
289 inc->sample_buffer[j / 8] |= (1 << (j % 8));
298 static char **parse_line(char *buf, struct context *inc, int max_columns)
300 const char *str, *remainder;
311 str = strstr(remainder, inc->delimiter->str);
313 while (str && max_columns) {
314 if (n >= inc->first_column) {
315 column = g_strndup(remainder, str - remainder);
316 list = g_slist_prepend(list, g_strstrip(column));
322 remainder = str + inc->delimiter->len;
323 str = strstr(remainder, inc->delimiter->str);
327 if (buf[0] && max_columns && n >= inc->first_column) {
328 column = g_strdup(remainder);
329 list = g_slist_prepend(list, g_strstrip(column));
333 if (!(columns = g_try_new(char *, k + 1)))
338 for (l = list; l; l = l->next)
339 columns[k--] = l->data;
346 static int parse_multi_columns(char **columns, struct context *inc)
351 /* Clear buffer in order to set bits only. */
352 memset(inc->sample_buffer, 0, (inc->num_channels + 7) >> 3);
354 for (i = 0; i < inc->num_channels; i++) {
356 if (column[0] == '1') {
357 inc->sample_buffer[i / 8] |= (1 << (i % 8));
358 } else if (!strlen(column)) {
359 sr_err("Column %zu in line %zu is empty.",
360 inc->first_channel + i, inc->line_number);
362 } else if (column[0] != '0') {
363 sr_err("Invalid value '%s' in column %zu in line %zu.",
364 column, inc->first_channel + i,
373 static int parse_single_column(const char *column, struct context *inc)
379 switch (inc->format) {
381 res = parse_binstr(column, inc);
384 res = parse_hexstr(column, inc);
387 res = parse_octstr(column, inc);
394 static int send_samples(const struct sr_dev_inst *sdi, uint8_t *buffer,
395 gsize buffer_size, gsize count)
397 struct sr_datafeed_packet packet;
398 struct sr_datafeed_logic logic;
402 packet.type = SR_DF_LOGIC;
403 packet.payload = &logic;
404 logic.unitsize = buffer_size;
405 logic.length = buffer_size;
408 for (i = 0; i < count; i++) {
409 res = sr_session_send(sdi, &packet);
417 static int init(struct sr_input *in, GHashTable *options)
422 in->sdi = g_malloc0(sizeof(struct sr_dev_inst));
423 in->priv = inc = g_malloc0(sizeof(struct context));
425 inc->single_column = g_variant_get_int32(g_hash_table_lookup(options, "single-column"));
426 inc->multi_column_mode = inc->single_column == 0;
428 inc->num_channels = g_variant_get_int32(g_hash_table_lookup(options, "numchannels"));
430 inc->delimiter = g_string_new(g_variant_get_string(
431 g_hash_table_lookup(options, "delimiter"), NULL));
432 if (inc->delimiter->len == 0) {
433 sr_err("Delimiter must be at least one character.");
437 s = g_variant_get_string(g_hash_table_lookup(options, "format"), NULL);
438 if (!g_ascii_strncasecmp(s, "bin", 3)) {
439 inc->format = FORMAT_BIN;
440 } else if (!g_ascii_strncasecmp(s, "hex", 3)) {
441 inc->format = FORMAT_HEX;
442 } else if (!g_ascii_strncasecmp(s, "oct", 3)) {
443 inc->format = FORMAT_OCT;
445 sr_err("Invalid format: '%s'", s);
449 inc->comment = g_string_new(g_variant_get_string(
450 g_hash_table_lookup(options, "comment"), NULL));
451 if (g_string_equal(inc->comment, inc->delimiter)) {
452 /* That's never going to work. Likely the result of the user
453 * setting the delimiter to ; -- the default comment. Clearing
454 * the comment setting will work in that case. */
455 g_string_truncate(inc->comment, 0);
458 inc->samplerate = g_variant_get_uint64(g_hash_table_lookup(options, "samplerate"));
460 inc->first_channel = g_variant_get_int32(g_hash_table_lookup(options, "first-channel"));
462 inc->header = g_variant_get_boolean(g_hash_table_lookup(options, "header"));
464 inc->start_line = g_variant_get_int32(g_hash_table_lookup(options, "startline"));
465 if (inc->start_line < 1) {
466 sr_err("Invalid start line %zu.", inc->start_line);
470 if (inc->multi_column_mode)
471 inc->first_column = inc->first_channel;
473 inc->first_column = inc->single_column;
475 if (!inc->multi_column_mode && !inc->num_channels) {
476 sr_err("Number of channels needs to be specified in single column mode.");
483 static const char *delim_set = "\r\n";
485 static const char *get_line_termination(GString *buf)
490 if (g_strstr_len(buf->str, buf->len, "\r\n"))
492 else if (memchr(buf->str, '\n', buf->len))
494 else if (memchr(buf->str, '\r', buf->len))
500 static int initial_parse(const struct sr_input *in, GString *buf)
503 GString *channel_name;
504 unsigned int num_columns, i;
505 size_t line_number, l;
507 char **lines, *line, **columns, *column;
514 lines = g_strsplit_set(buf->str, delim_set, 0);
515 for (l = 0; lines[l]; l++) {
518 if (inc->start_line > line_number) {
519 sr_spew("Line %zu skipped.", line_number);
522 if (line[0] == '\0') {
523 sr_spew("Blank line %zu skipped.", line_number);
526 strip_comment(line, inc->comment);
527 if (line[0] == '\0') {
528 sr_spew("Comment-only line %zu skipped.", line_number);
532 /* Reached first proper line. */
536 /* Not enough data for a proper line yet. */
542 * In order to determine the number of columns parse the current line
543 * without limiting the number of columns.
545 columns = parse_line(line, inc, -1);
547 sr_err("Error while parsing line %zu.", line_number);
551 num_columns = g_strv_length(columns);
553 /* Ensure that the first column is not out of bounds. */
555 sr_err("Column %u in line %zu is out of bounds.",
556 inc->first_column, line_number);
561 if (inc->multi_column_mode) {
563 * Detect the number of channels in multi column mode
564 * automatically if not specified.
566 if (!inc->num_channels) {
567 inc->num_channels = num_columns;
568 sr_dbg("Number of auto-detected channels: %u.",
573 * Ensure that the number of channels does not exceed the number
574 * of columns in multi column mode.
576 if (num_columns < inc->num_channels) {
577 sr_err("Not enough columns for desired number of channels in line %zu.",
584 channel_name = g_string_sized_new(64);
585 for (i = 0; i < inc->num_channels; i++) {
587 if (inc->header && inc->multi_column_mode && column[0] != '\0')
588 g_string_assign(channel_name, column);
590 g_string_printf(channel_name, "%u", i);
591 sr_channel_new(in->sdi, i, SR_CHANNEL_LOGIC, TRUE, channel_name->str);
593 g_string_free(channel_name, TRUE);
596 * Calculate the minimum buffer size to store the sample data of the
599 inc->sample_buffer_size = (inc->num_channels + 7) >> 3;
600 inc->sample_buffer = g_malloc(inc->sample_buffer_size);
611 * Gets called from initial_receive(), which runs until the end-of-line
612 * encoding of the input stream could get determined. Assumes that this
613 * routine receives enough buffered initial input data to either see the
614 * BOM when there is one, or that no BOM will follow when a text line
615 * termination sequence was seen. Silently drops the UTF-8 BOM sequence
616 * from the input buffer if one was seen. Does not care to protect
617 * against multiple execution or dropping the BOM multiple times --
618 * there should be at most one in the input stream.
620 static void initial_bom_check(const struct sr_input *in)
622 static const char *utf8_bom = "\xef\xbb\xbf";
624 if (in->buf->len < strlen(utf8_bom))
626 if (strncmp(in->buf->str, utf8_bom, strlen(utf8_bom)) != 0)
628 g_string_erase(in->buf, 0, strlen(utf8_bom));
631 static int initial_receive(const struct sr_input *in)
637 const char *termination;
639 initial_bom_check(in);
643 termination = get_line_termination(in->buf);
645 /* Don't have a full line yet. */
648 p = g_strrstr_len(in->buf->str, in->buf->len, termination);
650 /* Don't have a full line yet. */
652 len = p - in->buf->str - 1;
653 new_buf = g_string_new_len(in->buf->str, len);
654 g_string_append_c(new_buf, '\0');
656 inc->termination = g_strdup(termination);
658 if (in->buf->str[0] != '\0')
659 ret = initial_parse(in, new_buf);
663 g_string_free(new_buf, TRUE);
668 static int process_buffer(struct sr_input *in, gboolean is_eof)
670 struct sr_datafeed_packet packet;
671 struct sr_datafeed_meta meta;
672 struct sr_config *src;
676 int max_columns, ret, l;
677 char *p, **lines, *line, **columns;
681 std_session_send_df_header(in->sdi);
683 if (inc->samplerate) {
684 packet.type = SR_DF_META;
685 packet.payload = &meta;
686 samplerate = inc->samplerate;
687 src = sr_config_new(SR_CONF_SAMPLERATE, g_variant_new_uint64(samplerate));
688 meta.config = g_slist_append(NULL, src);
689 sr_session_send(in->sdi, &packet);
690 g_slist_free(meta.config);
697 /* Limit the number of columns to parse. */
698 if (inc->multi_column_mode)
699 max_columns = inc->num_channels;
704 * Consider empty input non-fatal. Keep accumulating input until
705 * at least one full text line has become available. Grab the
706 * maximum amount of accumulated data that consists of full text
707 * lines, and process what has been received so far, leaving not
708 * yet complete lines for the next invocation.
710 * Enforce that all previously buffered data gets processed in
711 * the "EOF" condition. Do not insist in the presence of the
712 * termination sequence for the last line (may often be missing
713 * on Windows). A present termination sequence will just result
714 * in the "execution of an empty line", and does not harm.
719 p = in->buf->str + in->buf->len;
721 p = g_strrstr_len(in->buf->str, in->buf->len, inc->termination);
725 p += strlen(inc->termination);
727 g_strstrip(in->buf->str);
730 lines = g_strsplit_set(in->buf->str, delim_set, 0);
731 for (l = 0; lines[l]; l++) {
734 if (line[0] == '\0') {
735 sr_spew("Blank line %zu skipped.", inc->line_number);
739 /* Remove trailing comment. */
740 strip_comment(line, inc->comment);
741 if (line[0] == '\0') {
742 sr_spew("Comment-only line %zu skipped.", inc->line_number);
746 /* Skip the header line, its content was used as the channel names. */
748 sr_spew("Header line %zu skipped.", inc->line_number);
753 columns = parse_line(line, inc, max_columns);
755 sr_err("Error while parsing line %zu.", inc->line_number);
758 num_columns = g_strv_length(columns);
760 sr_err("Column %u in line %zu is out of bounds.",
761 inc->first_column, inc->line_number);
766 * Ensure that the number of channels does not exceed the number
767 * of columns in multi column mode.
769 if (inc->multi_column_mode && num_columns < inc->num_channels) {
770 sr_err("Not enough columns for desired number of channels in line %zu.",
776 if (inc->multi_column_mode)
777 ret = parse_multi_columns(columns, inc);
779 ret = parse_single_column(columns[0], inc);
785 /* Send sample data to the session bus. */
786 ret = send_samples(in->sdi, inc->sample_buffer,
787 inc->sample_buffer_size, 1);
789 sr_err("Sending samples failed.");
795 g_string_erase(in->buf, 0, p - in->buf->str);
800 static int receive(struct sr_input *in, GString *buf)
805 g_string_append_len(in->buf, buf->str, buf->len);
808 if (!inc->termination) {
809 ret = initial_receive(in);
810 if (ret == SR_ERR_NA)
811 /* Not enough data yet. */
813 else if (ret != SR_OK)
816 /* sdi is ready, notify frontend. */
817 in->sdi_ready = TRUE;
821 ret = process_buffer(in, FALSE);
826 static int end(struct sr_input *in)
832 ret = process_buffer(in, TRUE);
838 std_session_send_df_end(in->sdi);
843 static void cleanup(struct sr_input *in)
850 g_string_free(inc->delimiter, TRUE);
853 g_string_free(inc->comment, TRUE);
855 g_free(inc->termination);
856 g_free(inc->sample_buffer);
859 static int reset(struct sr_input *in)
861 struct context *inc = in->priv;
864 inc->started = FALSE;
865 g_string_truncate(in->buf, 0);
870 static struct sr_option options[] = {
871 { "single-column", "Single column", "Enable/specify single column", NULL, NULL },
872 { "numchannels", "Max channels", "Number of channels", NULL, NULL },
873 { "delimiter", "Delimiter", "Column delimiter", NULL, NULL },
874 { "format", "Format", "Numeric format", NULL, NULL },
875 { "comment", "Comment", "Comment prefix character", NULL, NULL },
876 { "samplerate", "Samplerate", "Samplerate used during capture", NULL, NULL },
877 { "first-channel", "First channel", "Column number of first channel", NULL, NULL },
878 { "header", "Header", "Treat first line as header with channel names", NULL, NULL },
879 { "startline", "Start line", "Line number at which to start processing samples", NULL, NULL },
883 static const struct sr_option *get_options(void)
885 if (!options[0].def) {
886 options[0].def = g_variant_ref_sink(g_variant_new_int32(0));
887 options[1].def = g_variant_ref_sink(g_variant_new_int32(0));
888 options[2].def = g_variant_ref_sink(g_variant_new_string(","));
889 options[3].def = g_variant_ref_sink(g_variant_new_string("bin"));
890 options[4].def = g_variant_ref_sink(g_variant_new_string(";"));
891 options[5].def = g_variant_ref_sink(g_variant_new_uint64(0));
892 options[6].def = g_variant_ref_sink(g_variant_new_int32(0));
893 options[7].def = g_variant_ref_sink(g_variant_new_boolean(FALSE));
894 options[8].def = g_variant_ref_sink(g_variant_new_int32(1));
900 SR_PRIV struct sr_input_module input_csv = {
903 .desc = "Comma-separated values",
904 .exts = (const char*[]){"csv", NULL},
905 .options = get_options,