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. */
108 enum single_col_format {
117 /* Current selected samplerate. */
120 /* Number of 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 size_t 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.
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 size_t 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. */
161 enum single_col_format format;
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 %zu 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 %zu 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 %zu 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 %zu 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 %zu 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 %zu 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));
307 * @brief Splits a text line into a set of columns.
309 * @param[in] buf The input text line to split.
310 * @param[in] inc The input module's context.
311 * @param[in] max_cols The maximum column count, negative to get all of them.
313 * @returns An array of strings, representing the columns' text.
315 static char **parse_line(char *buf, struct context *inc, ssize_t max_cols)
317 const char *str, *remainder;
328 str = strstr(remainder, inc->delimiter->str);
330 while (str && max_cols) {
331 if (n >= inc->first_column) {
332 column = g_strndup(remainder, str - remainder);
333 list = g_slist_prepend(list, g_strstrip(column));
339 remainder = str + inc->delimiter->len;
340 str = strstr(remainder, inc->delimiter->str);
344 if (buf[0] && max_cols && n >= inc->first_column) {
345 column = g_strdup(remainder);
346 list = g_slist_prepend(list, g_strstrip(column));
350 if (!(columns = g_try_new(char *, k + 1)))
355 for (l = list; l; l = l->next)
356 columns[k--] = l->data;
363 static int parse_multi_columns(char **columns, struct context *inc)
368 /* Clear buffer in order to set bits only. */
369 memset(inc->sample_buffer, 0, inc->sample_unit_size);
371 for (i = 0; i < inc->num_channels; i++) {
373 if (column[0] == '1') {
374 inc->sample_buffer[i / 8] |= (1 << (i % 8));
375 } else if (!strlen(column)) {
376 sr_err("Column %zu in line %zu is empty.",
377 inc->first_channel + i, inc->line_number);
379 } else if (column[0] != '0') {
380 sr_err("Invalid value '%s' in column %zu in line %zu.",
381 column, inc->first_channel + i,
390 static int parse_single_column(const char *column, struct context *inc)
396 switch (inc->format) {
398 res = parse_binstr(column, inc);
401 res = parse_hexstr(column, inc);
404 res = parse_octstr(column, inc);
411 static int flush_samples(const struct sr_input *in)
414 struct sr_datafeed_packet packet;
415 struct sr_datafeed_logic logic;
419 if (!inc->datafeed_buf_fill)
422 memset(&packet, 0, sizeof(packet));
423 memset(&logic, 0, sizeof(logic));
424 packet.type = SR_DF_LOGIC;
425 packet.payload = &logic;
426 logic.unitsize = inc->sample_unit_size;
427 logic.length = inc->datafeed_buf_fill;
428 logic.data = inc->datafeed_buffer;
430 rc = sr_session_send(in->sdi, &packet);
434 inc->datafeed_buf_fill = 0;
438 static int queue_samples(const struct sr_input *in)
445 inc->datafeed_buf_fill += inc->sample_unit_size;
446 if (inc->datafeed_buf_fill == inc->datafeed_buf_size) {
447 rc = flush_samples(in);
451 inc->sample_buffer = &inc->datafeed_buffer[inc->datafeed_buf_fill];
455 static int init(struct sr_input *in, GHashTable *options)
460 in->sdi = g_malloc0(sizeof(struct sr_dev_inst));
461 in->priv = inc = g_malloc0(sizeof(struct context));
463 inc->single_column = g_variant_get_uint32(g_hash_table_lookup(options, "single-column"));
464 inc->multi_column_mode = inc->single_column == 0;
466 inc->num_channels = g_variant_get_uint32(g_hash_table_lookup(options, "numchannels"));
468 inc->delimiter = g_string_new(g_variant_get_string(
469 g_hash_table_lookup(options, "delimiter"), NULL));
470 if (inc->delimiter->len == 0) {
471 sr_err("Delimiter must be at least one character.");
475 s = g_variant_get_string(g_hash_table_lookup(options, "format"), NULL);
476 if (!g_ascii_strncasecmp(s, "bin", 3)) {
477 inc->format = FORMAT_BIN;
478 } else if (!g_ascii_strncasecmp(s, "hex", 3)) {
479 inc->format = FORMAT_HEX;
480 } else if (!g_ascii_strncasecmp(s, "oct", 3)) {
481 inc->format = FORMAT_OCT;
483 sr_err("Invalid format: '%s'", s);
487 inc->comment = g_string_new(g_variant_get_string(
488 g_hash_table_lookup(options, "comment"), NULL));
489 if (g_string_equal(inc->comment, inc->delimiter)) {
490 /* That's never going to work. Likely the result of the user
491 * setting the delimiter to ; -- the default comment. Clearing
492 * the comment setting will work in that case. */
493 g_string_truncate(inc->comment, 0);
496 inc->samplerate = g_variant_get_uint64(g_hash_table_lookup(options, "samplerate"));
498 inc->first_channel = g_variant_get_uint32(g_hash_table_lookup(options, "first-channel"));
500 inc->header = g_variant_get_boolean(g_hash_table_lookup(options, "header"));
502 inc->start_line = g_variant_get_uint32(g_hash_table_lookup(options, "startline"));
503 if (inc->start_line < 1) {
504 sr_err("Invalid start line %zu.", inc->start_line);
508 if (inc->multi_column_mode)
509 inc->first_column = inc->first_channel;
511 inc->first_column = inc->single_column;
513 if (!inc->multi_column_mode && !inc->num_channels) {
514 sr_err("Number of channels needs to be specified in single column mode.");
522 * Check the channel list for consistency across file re-import. See
523 * the VCD input module for more details and motivation.
526 static void keep_header_for_reread(const struct sr_input *in)
531 g_slist_free_full(inc->prev_sr_channels, sr_channel_free_cb);
532 inc->prev_sr_channels = in->sdi->channels;
533 in->sdi->channels = NULL;
536 static int check_header_in_reread(const struct sr_input *in)
545 if (!inc->prev_sr_channels)
548 if (sr_channel_lists_differ(inc->prev_sr_channels, in->sdi->channels)) {
549 sr_err("Channel list change not supported for file re-read.");
552 g_slist_free_full(in->sdi->channels, sr_channel_free_cb);
553 in->sdi->channels = inc->prev_sr_channels;
554 inc->prev_sr_channels = NULL;
559 static const char *delim_set = "\r\n";
561 static const char *get_line_termination(GString *buf)
566 if (g_strstr_len(buf->str, buf->len, "\r\n"))
568 else if (memchr(buf->str, '\n', buf->len))
570 else if (memchr(buf->str, '\r', buf->len))
576 static int initial_parse(const struct sr_input *in, GString *buf)
579 GString *channel_name;
580 size_t num_columns, i;
581 size_t line_number, l;
583 char **lines, *line, **columns, *column;
590 lines = g_strsplit_set(buf->str, delim_set, 0);
591 for (l = 0; lines[l]; l++) {
594 if (inc->start_line > line_number) {
595 sr_spew("Line %zu skipped.", line_number);
598 if (line[0] == '\0') {
599 sr_spew("Blank line %zu skipped.", line_number);
602 strip_comment(line, inc->comment);
603 if (line[0] == '\0') {
604 sr_spew("Comment-only line %zu skipped.", line_number);
608 /* Reached first proper line. */
612 /* Not enough data for a proper line yet. */
618 * In order to determine the number of columns parse the current line
619 * without limiting the number of columns.
621 columns = parse_line(line, inc, -1);
623 sr_err("Error while parsing line %zu.", line_number);
627 num_columns = g_strv_length(columns);
629 /* Ensure that the first column is not out of bounds. */
631 sr_err("Column %zu in line %zu is out of bounds.",
632 inc->first_column, line_number);
637 if (inc->multi_column_mode) {
639 * Detect the number of channels in multi column mode
640 * automatically if not specified.
642 if (!inc->num_channels) {
643 inc->num_channels = num_columns;
644 sr_dbg("Number of auto-detected channels: %zu.",
649 * Ensure that the number of channels does not exceed the number
650 * of columns in multi column mode.
652 if (num_columns < inc->num_channels) {
653 sr_err("Not enough columns for desired number of channels in line %zu.",
660 channel_name = g_string_sized_new(64);
661 for (i = 0; i < inc->num_channels; i++) {
663 if (inc->header && inc->multi_column_mode && column[0] != '\0')
664 g_string_assign(channel_name, column);
666 g_string_printf(channel_name, "%zu", i);
667 sr_channel_new(in->sdi, i, SR_CHANNEL_LOGIC, TRUE, channel_name->str);
669 g_string_free(channel_name, TRUE);
670 if (!check_header_in_reread(in)) {
676 * Calculate the minimum buffer size to store the set of samples
677 * of all channels (unit size). Determine a larger buffer size
678 * for datafeed submission that is a multiple of the unit size.
679 * Allocate the larger buffer, and have the "sample buffer" point
680 * to a location within that large buffer.
682 inc->sample_unit_size = (inc->num_channels + 7) / 8;
683 inc->datafeed_buf_size = CHUNK_SIZE;
684 inc->datafeed_buf_size *= inc->sample_unit_size;
685 inc->datafeed_buffer = g_malloc(inc->datafeed_buf_size);
686 inc->datafeed_buf_fill = 0;
687 inc->sample_buffer = &inc->datafeed_buffer[inc->datafeed_buf_fill];
698 * Gets called from initial_receive(), which runs until the end-of-line
699 * encoding of the input stream could get determined. Assumes that this
700 * routine receives enough buffered initial input data to either see the
701 * BOM when there is one, or that no BOM will follow when a text line
702 * termination sequence was seen. Silently drops the UTF-8 BOM sequence
703 * from the input buffer if one was seen. Does not care to protect
704 * against multiple execution or dropping the BOM multiple times --
705 * there should be at most one in the input stream.
707 static void initial_bom_check(const struct sr_input *in)
709 static const char *utf8_bom = "\xef\xbb\xbf";
711 if (in->buf->len < strlen(utf8_bom))
713 if (strncmp(in->buf->str, utf8_bom, strlen(utf8_bom)) != 0)
715 g_string_erase(in->buf, 0, strlen(utf8_bom));
718 static int initial_receive(const struct sr_input *in)
724 const char *termination;
726 initial_bom_check(in);
730 termination = get_line_termination(in->buf);
732 /* Don't have a full line yet. */
735 p = g_strrstr_len(in->buf->str, in->buf->len, termination);
737 /* Don't have a full line yet. */
739 len = p - in->buf->str - 1;
740 new_buf = g_string_new_len(in->buf->str, len);
741 g_string_append_c(new_buf, '\0');
743 inc->termination = g_strdup(termination);
745 if (in->buf->str[0] != '\0')
746 ret = initial_parse(in, new_buf);
750 g_string_free(new_buf, TRUE);
755 static int process_buffer(struct sr_input *in, gboolean is_eof)
757 struct sr_datafeed_packet packet;
758 struct sr_datafeed_meta meta;
759 struct sr_config *src;
763 size_t max_columns, l;
765 char *p, **lines, *line, **columns;
769 std_session_send_df_header(in->sdi);
771 if (inc->samplerate) {
772 packet.type = SR_DF_META;
773 packet.payload = &meta;
774 samplerate = inc->samplerate;
775 src = sr_config_new(SR_CONF_SAMPLERATE, g_variant_new_uint64(samplerate));
776 meta.config = g_slist_append(NULL, src);
777 sr_session_send(in->sdi, &packet);
778 g_slist_free(meta.config);
785 /* Limit the number of columns to parse. */
786 if (inc->multi_column_mode)
787 max_columns = inc->num_channels;
792 * Consider empty input non-fatal. Keep accumulating input until
793 * at least one full text line has become available. Grab the
794 * maximum amount of accumulated data that consists of full text
795 * lines, and process what has been received so far, leaving not
796 * yet complete lines for the next invocation.
798 * Enforce that all previously buffered data gets processed in
799 * the "EOF" condition. Do not insist in the presence of the
800 * termination sequence for the last line (may often be missing
801 * on Windows). A present termination sequence will just result
802 * in the "execution of an empty line", and does not harm.
807 p = in->buf->str + in->buf->len;
809 p = g_strrstr_len(in->buf->str, in->buf->len, inc->termination);
813 p += strlen(inc->termination);
815 g_strstrip(in->buf->str);
818 lines = g_strsplit_set(in->buf->str, delim_set, 0);
819 for (l = 0; lines[l]; l++) {
822 if (line[0] == '\0') {
823 sr_spew("Blank line %zu skipped.", inc->line_number);
827 /* Remove trailing comment. */
828 strip_comment(line, inc->comment);
829 if (line[0] == '\0') {
830 sr_spew("Comment-only line %zu skipped.", inc->line_number);
834 /* Skip the header line, its content was used as the channel names. */
836 sr_spew("Header line %zu skipped.", inc->line_number);
841 columns = parse_line(line, inc, max_columns);
843 sr_err("Error while parsing line %zu.", inc->line_number);
847 num_columns = g_strv_length(columns);
849 sr_err("Column %zu in line %zu is out of bounds.",
850 inc->first_column, inc->line_number);
856 * Ensure that the number of channels does not exceed the number
857 * of columns in multi column mode.
859 if (inc->multi_column_mode && num_columns < inc->num_channels) {
860 sr_err("Not enough columns for desired number of channels in line %zu.",
867 if (inc->multi_column_mode)
868 ret = parse_multi_columns(columns, inc);
870 ret = parse_single_column(columns[0], inc);
877 /* Send sample data to the session bus. */
878 ret = queue_samples(in);
880 sr_err("Sending samples failed.");
889 g_string_erase(in->buf, 0, p - in->buf->str);
894 static int receive(struct sr_input *in, GString *buf)
899 g_string_append_len(in->buf, buf->str, buf->len);
902 if (!inc->termination) {
903 ret = initial_receive(in);
904 if (ret == SR_ERR_NA)
905 /* Not enough data yet. */
907 else if (ret != SR_OK)
910 /* sdi is ready, notify frontend. */
911 in->sdi_ready = TRUE;
915 ret = process_buffer(in, FALSE);
920 static int end(struct sr_input *in)
926 ret = process_buffer(in, TRUE);
932 ret = flush_samples(in);
938 std_session_send_df_end(in->sdi);
943 static void cleanup(struct sr_input *in)
947 keep_header_for_reread(in);
951 g_free(inc->termination);
952 inc->termination = NULL;
953 g_free(inc->datafeed_buffer);
954 inc->datafeed_buffer = NULL;
957 static int reset(struct sr_input *in)
959 struct context *inc = in->priv;
962 inc->started = FALSE;
963 g_string_truncate(in->buf, 0);
981 static struct sr_option options[] = {
982 [OPT_SINGLE_COL] = { "single-column", "Single column", "Enable single-column mode, using the specified column (>= 1); 0: multi-col. mode", NULL, NULL },
983 [OPT_NUM_LOGIC] = { "numchannels", "Number of logic channels", "The number of (logic) channels (single-col. mode: number of bits beginning at 'first channel', LSB-first)", NULL, NULL },
984 [OPT_DELIM] = { "delimiter", "Column delimiter", "The column delimiter (>= 1 characters)", NULL, NULL },
985 [OPT_FORMAT] = { "format", "Data format (single-col. mode)", "The numeric format of the data (single-col. mode): bin, hex, oct", NULL, NULL },
986 [OPT_COMMENT] = { "comment", "Comment character(s)", "The comment prefix character(s)", NULL, NULL },
987 [OPT_RATE] = { "samplerate", "Samplerate (Hz)", "The sample rate (used during capture) in Hz", NULL, NULL },
988 [OPT_FIRST_LOGIC] = { "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 },
989 [OPT_HEADER] = { "header", "Interpret first line as header (multi-col. mode)", "Treat the first line as header with channel names (multi-col. mode)", NULL, NULL },
990 [OPT_START] = { "startline", "Start line", "The line number at which to start processing samples (>= 1)", NULL, NULL },
991 [OPT_MAX] = ALL_ZERO,
994 static const struct sr_option *get_options(void)
998 if (!options[0].def) {
999 options[OPT_SINGLE_COL].def = g_variant_ref_sink(g_variant_new_int32(0));
1000 options[OPT_NUM_LOGIC].def = g_variant_ref_sink(g_variant_new_int32(0));
1001 options[OPT_DELIM].def = g_variant_ref_sink(g_variant_new_string(","));
1002 options[OPT_FORMAT].def = g_variant_ref_sink(g_variant_new_string("bin"));
1004 l = g_slist_append(l, g_variant_ref_sink(g_variant_new_string("bin")));
1005 l = g_slist_append(l, g_variant_ref_sink(g_variant_new_string("hex")));
1006 l = g_slist_append(l, g_variant_ref_sink(g_variant_new_string("oct")));
1007 options[OPT_FORMAT].values = l;
1008 options[OPT_COMMENT].def = g_variant_ref_sink(g_variant_new_string(";"));
1009 options[OPT_RATE].def = g_variant_ref_sink(g_variant_new_uint64(0));
1010 options[OPT_FIRST_LOGIC].def = g_variant_ref_sink(g_variant_new_int32(0));
1011 options[OPT_HEADER].def = g_variant_ref_sink(g_variant_new_boolean(FALSE));
1012 options[OPT_START].def = g_variant_ref_sink(g_variant_new_int32(1));
1018 SR_PRIV struct sr_input_module input_csv = {
1021 .desc = "Comma-separated values",
1022 .exts = (const char*[]){"csv", NULL},
1023 .options = get_options,