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"
29 #define CHUNK_SIZE (4 * 1024 * 1024)
32 * The CSV input module has the following options:
34 * single-column: Specifies the column number which stores the sample data for
35 * single column mode and enables single column mode. Multi
36 * column mode is used if this parameter is omitted.
38 * numchannels: Specifies the number of channels to use. In multi column mode
39 * the number of channels are the number of columns and in single
40 * column mode the number of bits (LSB first) beginning at
43 * delimiter: Specifies the delimiter for columns. Must be at least one
44 * character. Comma is used as default delimiter.
46 * format: Specifies the format of the sample data in single column mode.
47 * Available formats are: 'bin', 'hex' and 'oct'. The binary
48 * format is used by default. This option has no effect in multi
51 * comment: Specifies the prefix character(s) for comments. No prefix
52 * characters are used by default which disables removing of
55 * samplerate: Samplerate which the sample data was captured with. Default
58 * first-channel: Column number of the first channel in multi column mode and
59 * position of the bit for the first channel in single column mode.
62 * header: Determines if the first line should be treated as header
63 * and used for channel names in multi column mode. Empty header
64 * names will be replaced by the channel number. If enabled in
65 * single column mode the first line will be skipped. Usage of
66 * header is disabled by default.
68 * startline: Line number to start processing sample data. Must be greater
69 * than 0. The default line number to start processing is 1.
75 * - Determine how the text line handling can get improved, regarding
76 * all of robustness and flexibility and correctness.
77 * - The current implementation splits on "any run of CR and LF". Which
78 * translates to: Line numbers are wrong in the presence of empty
79 * lines in the input stream. See below for an (expensive) fix.
80 * - Dropping support for CR style end-of-line markers could improve
81 * the situation a lot. Code could search for and split on LF, and
82 * trim optional trailing CR. This would result in proper support
83 * for CRLF (Windows) as well as LF (Unix), and allow for correct
85 * - When support for CR-only line termination cannot get dropped,
86 * then the current implementation is inappropriate. Currently the
87 * input stream is scanned for the first occurance of either of the
88 * supported termination styles (which is good). For the remaining
89 * session a consistent encoding of the text lines is assumed (which
91 * - When line numbers need to be correct and reliable, _and_ the full
92 * set of previously supported line termination sequences are required,
93 * and potentially more are to get added for improved compatibility
94 * with more platforms or generators, then the current approach of
95 * splitting on runs of termination characters needs to get replaced,
96 * by the more expensive approach to scan for and count the initially
97 * determined termination sequence.
99 * - Add support for analog input data? (optional)
100 * - Needs a syntax first for user specs which channels (columns) are
101 * logic and which are analog. May need heuristics(?) to guess from
102 * input data in the absence of user provided specs.
105 /* Single column formats. */
115 /* Current selected samplerate. */
118 /* Number of channels. */
119 unsigned int num_channels;
121 /* Column delimiter character(s). */
124 /* Comment prefix character(s). */
127 /* Termination character(s) used in current stream. */
130 /* Determines if sample data is stored in multiple columns. */
131 gboolean multi_column_mode;
133 /* Column number of the sample data in single column mode. */
134 unsigned int single_column;
137 * Number of the first column to parse. Equivalent to the number of the
138 * first channel in multi column mode and the single column number in
139 * single column mode.
141 unsigned int first_column;
144 * Column number of the first channel in multi column mode and position of
145 * the bit for the first channel in single column mode.
147 unsigned int first_channel;
149 /* Line number to start processing. */
153 * Determines if the first line should be treated as header and used for
154 * channel names in multi column mode.
158 /* Format sample data is stored in single column mode. */
161 size_t sample_unit_size; /**!< Byte count for a single sample. */
162 uint8_t *sample_buffer; /**!< Buffer for a single sample. */
164 uint8_t *datafeed_buffer; /**!< Queue for datafeed submission. */
165 size_t datafeed_buf_size;
166 size_t datafeed_buf_fill;
168 /* Current line number. */
171 /* List of previously created sigrok channels. */
172 GSList *prev_sr_channels;
175 static void strip_comment(char *buf, const GString *prefix)
182 if ((ptr = strstr(buf, prefix->str)))
186 static int parse_binstr(const char *str, struct context *inc)
190 length = strlen(str);
193 sr_err("Column %u in line %zu is empty.", inc->single_column,
198 /* Clear buffer in order to set bits only. */
199 memset(inc->sample_buffer, 0, inc->sample_unit_size);
201 i = inc->first_channel;
203 for (j = 0; i < length && j < inc->num_channels; i++, j++) {
204 if (str[length - i - 1] == '1') {
205 inc->sample_buffer[j / 8] |= (1 << (j % 8));
206 } else if (str[length - i - 1] != '0') {
207 sr_err("Invalid value '%s' in column %u in line %zu.",
208 str, inc->single_column, inc->line_number);
216 static int parse_hexstr(const char *str, struct context *inc)
218 gsize i, j, k, length;
222 length = strlen(str);
225 sr_err("Column %u in line %zu is empty.", inc->single_column,
230 /* Clear buffer in order to set bits only. */
231 memset(inc->sample_buffer, 0, inc->sample_unit_size);
233 /* Calculate the position of the first hexadecimal digit. */
234 i = inc->first_channel / 4;
236 for (j = 0; i < length && j < inc->num_channels; i++) {
237 c = str[length - i - 1];
239 if (!g_ascii_isxdigit(c)) {
240 sr_err("Invalid value '%s' in column %u in line %zu.",
241 str, inc->single_column, inc->line_number);
245 value = g_ascii_xdigit_value(c);
247 k = (inc->first_channel + j) % 4;
249 for (; j < inc->num_channels && k < 4; k++) {
250 if (value & (1 << k))
251 inc->sample_buffer[j / 8] |= (1 << (j % 8));
260 static int parse_octstr(const char *str, struct context *inc)
262 gsize i, j, k, length;
266 length = strlen(str);
269 sr_err("Column %u in line %zu is empty.", inc->single_column,
274 /* Clear buffer in order to set bits only. */
275 memset(inc->sample_buffer, 0, inc->sample_unit_size);
277 /* Calculate the position of the first octal digit. */
278 i = inc->first_channel / 3;
280 for (j = 0; i < length && j < inc->num_channels; i++) {
281 c = str[length - i - 1];
283 if (c < '0' || c > '7') {
284 sr_err("Invalid value '%s' in column %u in line %zu.",
285 str, inc->single_column, inc->line_number);
289 value = g_ascii_xdigit_value(c);
291 k = (inc->first_channel + j) % 3;
293 for (; j < inc->num_channels && k < 3; k++) {
294 if (value & (1 << k))
295 inc->sample_buffer[j / 8] |= (1 << (j % 8));
304 static char **parse_line(char *buf, struct context *inc, int max_columns)
306 const char *str, *remainder;
317 str = strstr(remainder, inc->delimiter->str);
319 while (str && max_columns) {
320 if (n >= inc->first_column) {
321 column = g_strndup(remainder, str - remainder);
322 list = g_slist_prepend(list, g_strstrip(column));
328 remainder = str + inc->delimiter->len;
329 str = strstr(remainder, inc->delimiter->str);
333 if (buf[0] && max_columns && n >= inc->first_column) {
334 column = g_strdup(remainder);
335 list = g_slist_prepend(list, g_strstrip(column));
339 if (!(columns = g_try_new(char *, k + 1)))
344 for (l = list; l; l = l->next)
345 columns[k--] = l->data;
352 static int parse_multi_columns(char **columns, struct context *inc)
357 /* Clear buffer in order to set bits only. */
358 memset(inc->sample_buffer, 0, inc->sample_unit_size);
360 for (i = 0; i < inc->num_channels; i++) {
362 if (column[0] == '1') {
363 inc->sample_buffer[i / 8] |= (1 << (i % 8));
364 } else if (!strlen(column)) {
365 sr_err("Column %zu in line %zu is empty.",
366 inc->first_channel + i, inc->line_number);
368 } else if (column[0] != '0') {
369 sr_err("Invalid value '%s' in column %zu in line %zu.",
370 column, inc->first_channel + i,
379 static int parse_single_column(const char *column, struct context *inc)
385 switch (inc->format) {
387 res = parse_binstr(column, inc);
390 res = parse_hexstr(column, inc);
393 res = parse_octstr(column, inc);
400 static int flush_samples(const struct sr_input *in)
403 struct sr_datafeed_packet packet;
404 struct sr_datafeed_logic logic;
408 if (!inc->datafeed_buf_fill)
411 memset(&packet, 0, sizeof(packet));
412 memset(&logic, 0, sizeof(logic));
413 packet.type = SR_DF_LOGIC;
414 packet.payload = &logic;
415 logic.unitsize = inc->sample_unit_size;
416 logic.length = inc->datafeed_buf_fill;
417 logic.data = inc->datafeed_buffer;
419 rc = sr_session_send(in->sdi, &packet);
423 inc->datafeed_buf_fill = 0;
427 static int queue_samples(const struct sr_input *in)
434 inc->datafeed_buf_fill += inc->sample_unit_size;
435 if (inc->datafeed_buf_fill == inc->datafeed_buf_size) {
436 rc = flush_samples(in);
440 inc->sample_buffer = &inc->datafeed_buffer[inc->datafeed_buf_fill];
444 static int init(struct sr_input *in, GHashTable *options)
449 in->sdi = g_malloc0(sizeof(struct sr_dev_inst));
450 in->priv = inc = g_malloc0(sizeof(struct context));
452 inc->single_column = g_variant_get_int32(g_hash_table_lookup(options, "single-column"));
453 inc->multi_column_mode = inc->single_column == 0;
455 inc->num_channels = g_variant_get_int32(g_hash_table_lookup(options, "numchannels"));
457 inc->delimiter = g_string_new(g_variant_get_string(
458 g_hash_table_lookup(options, "delimiter"), NULL));
459 if (inc->delimiter->len == 0) {
460 sr_err("Delimiter must be at least one character.");
464 s = g_variant_get_string(g_hash_table_lookup(options, "format"), NULL);
465 if (!g_ascii_strncasecmp(s, "bin", 3)) {
466 inc->format = FORMAT_BIN;
467 } else if (!g_ascii_strncasecmp(s, "hex", 3)) {
468 inc->format = FORMAT_HEX;
469 } else if (!g_ascii_strncasecmp(s, "oct", 3)) {
470 inc->format = FORMAT_OCT;
472 sr_err("Invalid format: '%s'", s);
476 inc->comment = g_string_new(g_variant_get_string(
477 g_hash_table_lookup(options, "comment"), NULL));
478 if (g_string_equal(inc->comment, inc->delimiter)) {
479 /* That's never going to work. Likely the result of the user
480 * setting the delimiter to ; -- the default comment. Clearing
481 * the comment setting will work in that case. */
482 g_string_truncate(inc->comment, 0);
485 inc->samplerate = g_variant_get_uint64(g_hash_table_lookup(options, "samplerate"));
487 inc->first_channel = g_variant_get_int32(g_hash_table_lookup(options, "first-channel"));
489 inc->header = g_variant_get_boolean(g_hash_table_lookup(options, "header"));
491 inc->start_line = g_variant_get_int32(g_hash_table_lookup(options, "startline"));
492 if (inc->start_line < 1) {
493 sr_err("Invalid start line %zu.", inc->start_line);
497 if (inc->multi_column_mode)
498 inc->first_column = inc->first_channel;
500 inc->first_column = inc->single_column;
502 if (!inc->multi_column_mode && !inc->num_channels) {
503 sr_err("Number of channels needs to be specified in single column mode.");
511 * Check the channel list for consistency across file re-import. See
512 * the VCD input module for more details and motivation.
515 static void keep_header_for_reread(const struct sr_input *in)
520 g_slist_free_full(inc->prev_sr_channels, sr_channel_free_cb);
521 inc->prev_sr_channels = in->sdi->channels;
522 in->sdi->channels = NULL;
525 static int check_header_in_reread(const struct sr_input *in)
534 if (!inc->prev_sr_channels)
537 if (sr_channel_lists_differ(inc->prev_sr_channels, in->sdi->channels)) {
538 sr_err("Channel list change not supported for file re-read.");
541 g_slist_free_full(in->sdi->channels, sr_channel_free_cb);
542 in->sdi->channels = inc->prev_sr_channels;
543 inc->prev_sr_channels = NULL;
548 static const char *delim_set = "\r\n";
550 static const char *get_line_termination(GString *buf)
555 if (g_strstr_len(buf->str, buf->len, "\r\n"))
557 else if (memchr(buf->str, '\n', buf->len))
559 else if (memchr(buf->str, '\r', buf->len))
565 static int initial_parse(const struct sr_input *in, GString *buf)
568 GString *channel_name;
569 unsigned int num_columns, i;
570 size_t line_number, l;
572 char **lines, *line, **columns, *column;
579 lines = g_strsplit_set(buf->str, delim_set, 0);
580 for (l = 0; lines[l]; l++) {
583 if (inc->start_line > line_number) {
584 sr_spew("Line %zu skipped.", line_number);
587 if (line[0] == '\0') {
588 sr_spew("Blank line %zu skipped.", line_number);
591 strip_comment(line, inc->comment);
592 if (line[0] == '\0') {
593 sr_spew("Comment-only line %zu skipped.", line_number);
597 /* Reached first proper line. */
601 /* Not enough data for a proper line yet. */
607 * In order to determine the number of columns parse the current line
608 * without limiting the number of columns.
610 columns = parse_line(line, inc, -1);
612 sr_err("Error while parsing line %zu.", line_number);
616 num_columns = g_strv_length(columns);
618 /* Ensure that the first column is not out of bounds. */
620 sr_err("Column %u in line %zu is out of bounds.",
621 inc->first_column, line_number);
626 if (inc->multi_column_mode) {
628 * Detect the number of channels in multi column mode
629 * automatically if not specified.
631 if (!inc->num_channels) {
632 inc->num_channels = num_columns;
633 sr_dbg("Number of auto-detected channels: %u.",
638 * Ensure that the number of channels does not exceed the number
639 * of columns in multi column mode.
641 if (num_columns < inc->num_channels) {
642 sr_err("Not enough columns for desired number of channels in line %zu.",
649 channel_name = g_string_sized_new(64);
650 for (i = 0; i < inc->num_channels; i++) {
652 if (inc->header && inc->multi_column_mode && column[0] != '\0')
653 g_string_assign(channel_name, column);
655 g_string_printf(channel_name, "%u", i);
656 sr_channel_new(in->sdi, i, SR_CHANNEL_LOGIC, TRUE, channel_name->str);
658 g_string_free(channel_name, TRUE);
659 if (!check_header_in_reread(in)) {
665 * Calculate the minimum buffer size to store the set of samples
666 * of all channels (unit size). Determine a larger buffer size
667 * for datafeed submission that is a multiple of the unit size.
668 * Allocate the larger buffer, and have the "sample buffer" point
669 * to a location within that large buffer.
671 inc->sample_unit_size = (inc->num_channels + 7) / 8;
672 inc->datafeed_buf_size = CHUNK_SIZE;
673 inc->datafeed_buf_size *= inc->sample_unit_size;
674 inc->datafeed_buffer = g_malloc(inc->datafeed_buf_size);
675 inc->datafeed_buf_fill = 0;
676 inc->sample_buffer = &inc->datafeed_buffer[inc->datafeed_buf_fill];
687 * Gets called from initial_receive(), which runs until the end-of-line
688 * encoding of the input stream could get determined. Assumes that this
689 * routine receives enough buffered initial input data to either see the
690 * BOM when there is one, or that no BOM will follow when a text line
691 * termination sequence was seen. Silently drops the UTF-8 BOM sequence
692 * from the input buffer if one was seen. Does not care to protect
693 * against multiple execution or dropping the BOM multiple times --
694 * there should be at most one in the input stream.
696 static void initial_bom_check(const struct sr_input *in)
698 static const char *utf8_bom = "\xef\xbb\xbf";
700 if (in->buf->len < strlen(utf8_bom))
702 if (strncmp(in->buf->str, utf8_bom, strlen(utf8_bom)) != 0)
704 g_string_erase(in->buf, 0, strlen(utf8_bom));
707 static int initial_receive(const struct sr_input *in)
713 const char *termination;
715 initial_bom_check(in);
719 termination = get_line_termination(in->buf);
721 /* Don't have a full line yet. */
724 p = g_strrstr_len(in->buf->str, in->buf->len, termination);
726 /* Don't have a full line yet. */
728 len = p - in->buf->str - 1;
729 new_buf = g_string_new_len(in->buf->str, len);
730 g_string_append_c(new_buf, '\0');
732 inc->termination = g_strdup(termination);
734 if (in->buf->str[0] != '\0')
735 ret = initial_parse(in, new_buf);
739 g_string_free(new_buf, TRUE);
744 static int process_buffer(struct sr_input *in, gboolean is_eof)
746 struct sr_datafeed_packet packet;
747 struct sr_datafeed_meta meta;
748 struct sr_config *src;
752 int max_columns, ret, l;
753 char *p, **lines, *line, **columns;
757 std_session_send_df_header(in->sdi);
759 if (inc->samplerate) {
760 packet.type = SR_DF_META;
761 packet.payload = &meta;
762 samplerate = inc->samplerate;
763 src = sr_config_new(SR_CONF_SAMPLERATE, g_variant_new_uint64(samplerate));
764 meta.config = g_slist_append(NULL, src);
765 sr_session_send(in->sdi, &packet);
766 g_slist_free(meta.config);
773 /* Limit the number of columns to parse. */
774 if (inc->multi_column_mode)
775 max_columns = inc->num_channels;
780 * Consider empty input non-fatal. Keep accumulating input until
781 * at least one full text line has become available. Grab the
782 * maximum amount of accumulated data that consists of full text
783 * lines, and process what has been received so far, leaving not
784 * yet complete lines for the next invocation.
786 * Enforce that all previously buffered data gets processed in
787 * the "EOF" condition. Do not insist in the presence of the
788 * termination sequence for the last line (may often be missing
789 * on Windows). A present termination sequence will just result
790 * in the "execution of an empty line", and does not harm.
795 p = in->buf->str + in->buf->len;
797 p = g_strrstr_len(in->buf->str, in->buf->len, inc->termination);
801 p += strlen(inc->termination);
803 g_strstrip(in->buf->str);
806 lines = g_strsplit_set(in->buf->str, delim_set, 0);
807 for (l = 0; lines[l]; l++) {
810 if (line[0] == '\0') {
811 sr_spew("Blank line %zu skipped.", inc->line_number);
815 /* Remove trailing comment. */
816 strip_comment(line, inc->comment);
817 if (line[0] == '\0') {
818 sr_spew("Comment-only line %zu skipped.", inc->line_number);
822 /* Skip the header line, its content was used as the channel names. */
824 sr_spew("Header line %zu skipped.", inc->line_number);
829 columns = parse_line(line, inc, max_columns);
831 sr_err("Error while parsing line %zu.", inc->line_number);
835 num_columns = g_strv_length(columns);
837 sr_err("Column %u in line %zu is out of bounds.",
838 inc->first_column, inc->line_number);
844 * Ensure that the number of channels does not exceed the number
845 * of columns in multi column mode.
847 if (inc->multi_column_mode && num_columns < inc->num_channels) {
848 sr_err("Not enough columns for desired number of channels in line %zu.",
855 if (inc->multi_column_mode)
856 ret = parse_multi_columns(columns, inc);
858 ret = parse_single_column(columns[0], inc);
865 /* Send sample data to the session bus. */
866 ret = queue_samples(in);
868 sr_err("Sending samples failed.");
877 g_string_erase(in->buf, 0, p - in->buf->str);
882 static int receive(struct sr_input *in, GString *buf)
887 g_string_append_len(in->buf, buf->str, buf->len);
890 if (!inc->termination) {
891 ret = initial_receive(in);
892 if (ret == SR_ERR_NA)
893 /* Not enough data yet. */
895 else if (ret != SR_OK)
898 /* sdi is ready, notify frontend. */
899 in->sdi_ready = TRUE;
903 ret = process_buffer(in, FALSE);
908 static int end(struct sr_input *in)
914 ret = process_buffer(in, TRUE);
920 ret = flush_samples(in);
926 std_session_send_df_end(in->sdi);
931 static void cleanup(struct sr_input *in)
935 keep_header_for_reread(in);
939 g_free(inc->termination);
940 inc->termination = NULL;
941 g_free(inc->datafeed_buffer);
942 inc->datafeed_buffer = NULL;
945 static int reset(struct sr_input *in)
947 struct context *inc = in->priv;
950 inc->started = FALSE;
951 g_string_truncate(in->buf, 0);
956 static struct sr_option options[] = {
957 { "single-column", "Single column", "Enable single-column mode, using the specified column (>= 1); 0: multi-col. mode", NULL, NULL },
958 { "numchannels", "Number of logic channels", "The number of (logic) channels (single-col. mode: number of bits beginning at 'first channel', LSB-first)", NULL, NULL },
959 { "delimiter", "Column delimiter", "The column delimiter (>= 1 characters)", NULL, NULL },
960 { "format", "Data format (single-col. mode)", "The numeric format of the data (single-col. mode): bin, hex, oct", NULL, NULL },
961 { "comment", "Comment character(s)", "The comment prefix character(s)", NULL, NULL },
962 { "samplerate", "Samplerate (Hz)", "The sample rate (used during capture) in Hz", NULL, NULL },
963 { "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 },
964 { "header", "Interpret first line as header (multi-col. mode)", "Treat the first line as header with channel names (multi-col. mode)", NULL, NULL },
965 { "startline", "Start line", "The line number at which to start processing samples (>= 1)", NULL, NULL },
969 static const struct sr_option *get_options(void)
973 if (!options[0].def) {
974 options[0].def = g_variant_ref_sink(g_variant_new_int32(0));
975 options[1].def = g_variant_ref_sink(g_variant_new_int32(0));
976 options[2].def = g_variant_ref_sink(g_variant_new_string(","));
977 options[3].def = g_variant_ref_sink(g_variant_new_string("bin"));
979 l = g_slist_append(l, g_variant_ref_sink(g_variant_new_string("bin")));
980 l = g_slist_append(l, g_variant_ref_sink(g_variant_new_string("hex")));
981 l = g_slist_append(l, g_variant_ref_sink(g_variant_new_string("oct")));
982 options[3].values = l;
983 options[4].def = g_variant_ref_sink(g_variant_new_string(";"));
984 options[5].def = g_variant_ref_sink(g_variant_new_uint64(0));
985 options[6].def = g_variant_ref_sink(g_variant_new_int32(0));
986 options[7].def = g_variant_ref_sink(g_variant_new_boolean(FALSE));
987 options[8].def = g_variant_ref_sink(g_variant_new_int32(1));
993 SR_PRIV struct sr_input_module input_csv = {
996 .desc = "Comma-separated values",
997 .exts = (const char*[]){"csv", NULL},
998 .options = get_options,