2 * This file is part of the libsigrok project.
4 * Copyright (C) 2013 Marc Schink <sigrok-dev@marcschink.de>
5 * Copyright (C) 2019 Gerhard Sittig <gerhard.sittig@gmx.net>
7 * This program is free software: you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License as published by
9 * the Free Software Foundation, either version 3 of the License, or
10 * (at your option) any later version.
12 * This program is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 * GNU General Public License for more details.
17 * You should have received a copy of the GNU General Public License
18 * along with this program. If not, see <http://www.gnu.org/licenses/>.
27 #include <libsigrok/libsigrok.h>
28 #include "libsigrok-internal.h"
29 #include "scpi.h" /* String un-quote for channel name from header line. */
31 #define LOG_PREFIX "input/csv"
33 #define CHUNK_SIZE (4 * 1024 * 1024)
36 * The CSV input module has the following options:
38 * single-column: Specifies the column number which stores the sample data for
39 * single column mode and enables single column mode. Multi
40 * column mode is used if this parameter is omitted.
42 * numchannels: Specifies the number of channels to use. In multi column mode
43 * the number of channels are the number of columns and in single
44 * column mode the number of bits (LSB first) beginning at
47 * delimiter: Specifies the delimiter for columns. Must be at least one
48 * character. Comma is used as default delimiter.
50 * format: Specifies the format of the sample data in single column mode.
51 * Available formats are: 'bin', 'hex' and 'oct'. The binary
52 * format is used by default. This option has no effect in multi
55 * comment: Specifies the prefix character(s) for comments. No prefix
56 * characters are used by default which disables removing of
59 * samplerate: Samplerate which the sample data was captured with. Default
62 * first-channel: Column number of the first channel in multi column mode and
63 * position of the bit for the first channel in single column mode.
66 * header: Determines if the first line should be treated as header
67 * and used for channel names in multi column mode. Empty header
68 * names will be replaced by the channel number. If enabled in
69 * single column mode the first line will be skipped. Usage of
70 * header is disabled by default.
72 * startline: Line number to start processing sample data. Must be greater
73 * than 0. The default line number to start processing is 1.
79 * - Determine how the text line handling can get improved, regarding
80 * all of robustness and flexibility and correctness.
81 * - The current implementation splits on "any run of CR and LF". Which
82 * translates to: Line numbers are wrong in the presence of empty
83 * lines in the input stream. See below for an (expensive) fix.
84 * - Dropping support for CR style end-of-line markers could improve
85 * the situation a lot. Code could search for and split on LF, and
86 * trim optional trailing CR. This would result in proper support
87 * for CRLF (Windows) as well as LF (Unix), and allow for correct
89 * - When support for CR-only line termination cannot get dropped,
90 * then the current implementation is inappropriate. Currently the
91 * input stream is scanned for the first occurance of either of the
92 * supported termination styles (which is good). For the remaining
93 * session a consistent encoding of the text lines is assumed (which
95 * - When line numbers need to be correct and reliable, _and_ the full
96 * set of previously supported line termination sequences are required,
97 * and potentially more are to get added for improved compatibility
98 * with more platforms or generators, then the current approach of
99 * splitting on runs of termination characters needs to get replaced,
100 * by the more expensive approach to scan for and count the initially
101 * determined termination sequence.
103 * - Add support for analog input data? (optional)
104 * - Needs a syntax first for user specs which channels (columns) are
105 * logic and which are analog. May need heuristics(?) to guess from
106 * input data in the absence of user provided specs.
109 /* Single column formats. */
110 enum single_col_format {
111 FORMAT_NONE, /* Ignore this column. */
112 FORMAT_BIN, /* Bin digits for a set of bits (or just one bit). */
113 FORMAT_HEX, /* Hex digits for a set of bits. */
114 FORMAT_OCT, /* Oct digits for a set of bits. */
117 static const char *col_format_text[] = {
118 [FORMAT_NONE] = "unknown",
119 [FORMAT_BIN] = "binary",
120 [FORMAT_HEX] = "hexadecimal",
121 [FORMAT_OCT] = "octal",
124 struct column_details {
126 enum single_col_format text_format;
127 size_t channel_offset;
128 size_t channel_count;
134 /* Current selected samplerate. */
136 gboolean samplerate_sent;
138 /* Number of logic channels. */
139 size_t logic_channels;
141 /* Column delimiter (actually separator), comment leader, EOL sequence. */
147 * Determines if sample data is stored in multiple columns,
148 * which column to start at, and how many columns to expect.
150 gboolean multi_column_mode;
152 size_t column_want_count;
153 /* Parameters how to process the columns. */
154 struct column_details *column_details;
156 /* Line number to start processing. */
160 * Determines if the first line should be treated as header and used for
161 * channel names in multi column mode.
164 gboolean header_seen;
166 size_t sample_unit_size; /**!< Byte count for a single sample. */
167 uint8_t *sample_buffer; /**!< Buffer for a single sample. */
169 uint8_t *datafeed_buffer; /**!< Queue for datafeed submission. */
170 size_t datafeed_buf_size;
171 size_t datafeed_buf_fill;
173 /* Current line number. */
176 /* List of previously created sigrok channels. */
177 GSList *prev_sr_channels;
181 * Primitive operations to handle sample sets:
182 * - Keep a buffer for datafeed submission, capable of holding many
183 * samples (reduces call overhead, improves throughput).
184 * - Have a "current sample set" pointer reference one position in that
185 * large samples buffer.
186 * - Clear the current sample set before text line inspection, then set
187 * the bits which are found active in the current line of text input.
188 * Phrase the API such that call sites can be kept simple. Advance to
189 * the next sample set between lines, flush the larger buffer as needed
190 * (when it is full, or upon EOF).
193 static void clear_logic_samples(struct context *inc)
195 inc->sample_buffer = &inc->datafeed_buffer[inc->datafeed_buf_fill];
196 memset(inc->sample_buffer, 0, inc->sample_unit_size);
199 static void set_logic_level(struct context *inc, size_t ch_idx, int on)
201 size_t byte_idx, bit_idx;
204 if (ch_idx >= inc->logic_channels)
209 byte_idx = ch_idx / 8;
210 bit_idx = ch_idx % 8;
211 bit_mask = 1 << bit_idx;
212 inc->sample_buffer[byte_idx] |= bit_mask;
215 static int flush_logic_samples(const struct sr_input *in)
218 struct sr_datafeed_packet packet;
219 struct sr_datafeed_meta meta;
220 struct sr_config *src;
222 struct sr_datafeed_logic logic;
226 if (!inc->datafeed_buf_fill)
229 if (inc->samplerate && !inc->samplerate_sent) {
230 packet.type = SR_DF_META;
231 packet.payload = &meta;
232 samplerate = inc->samplerate;
233 src = sr_config_new(SR_CONF_SAMPLERATE, g_variant_new_uint64(samplerate));
234 meta.config = g_slist_append(NULL, src);
235 sr_session_send(in->sdi, &packet);
236 g_slist_free(meta.config);
238 inc->samplerate_sent = TRUE;
241 memset(&packet, 0, sizeof(packet));
242 memset(&logic, 0, sizeof(logic));
243 packet.type = SR_DF_LOGIC;
244 packet.payload = &logic;
245 logic.unitsize = inc->sample_unit_size;
246 logic.length = inc->datafeed_buf_fill;
247 logic.data = inc->datafeed_buffer;
249 rc = sr_session_send(in->sdi, &packet);
253 inc->datafeed_buf_fill = 0;
257 static int queue_logic_samples(const struct sr_input *in)
263 if (!inc->logic_channels)
266 inc->datafeed_buf_fill += inc->sample_unit_size;
267 if (inc->datafeed_buf_fill == inc->datafeed_buf_size) {
268 rc = flush_logic_samples(in);
275 /* Helpers for "column processing". */
277 static int split_column_format(const char *spec,
278 size_t *column_count, enum single_col_format *format, size_t *bit_count)
281 char *endp, format_char;
282 enum single_col_format format_code;
287 /* Get the (optional, decimal, default 1) column count. */
289 count = strtoul(spec, &endp, 10);
295 *column_count = count;
298 /* Get the (mandatory, single letter) type spec (-/xob/l). */
299 format_char = *spec++;
300 switch (format_char) {
301 case '-': /* Might conflict with number-parsing. */
304 format_code = FORMAT_NONE;
307 format_code = FORMAT_HEX;
310 format_code = FORMAT_OCT;
314 format_code = FORMAT_BIN;
316 default: /* includes NUL */
320 *format = format_code;
322 /* Get the (optional, decimal, default 1) bit count. */
324 count = strtoul(spec, &endp, 10);
329 if (format_char == '-')
331 if (format_char == 'l')
337 /* Input spec must have been exhausted. */
344 static int make_column_details_from_format(struct context *inc,
345 const char *column_format)
347 char **formats, *format;
348 size_t format_count, column_count, bit_count;
349 size_t format_idx, c, b, column_idx, channel_idx;
350 enum single_col_format f;
351 struct column_details *detail;
355 * Default to "all single-bit logic in each column" (which is
356 * the former multi-column mode).
358 if (!column_format || !*column_format) {
359 sr_dbg("Missing columns format, assuming multi-column mode.");
360 column_format = "0l";
363 /* Split the input spec, count involved columns and bits. */
364 formats = g_strsplit(column_format, ",", 0);
366 sr_err("Cannot parse columns format %s (comma split).", column_format);
369 format_count = g_strv_length(formats);
371 sr_err("Cannot parse columns format %s (field count).", column_format);
375 column_count = bit_count = 0;
376 for (format_idx = 0; format_idx < format_count; format_idx++) {
377 format = formats[format_idx];
378 ret = split_column_format(format, &c, &f, &b);
379 sr_dbg("fmt %s -> %zu cols, %s fmt, %zu bits, rc %d", format, c, col_format_text[f], b, ret);
381 sr_err("Cannot parse columns format %s (field split, %s).", column_format, format);
388 sr_dbg("Column format %s -> %zu columns, %zu logic channels.",
389 column_format, column_count, bit_count);
391 /* Allocate and fill in "column processing" details. */
392 inc->column_want_count = column_count;
393 inc->column_details = g_malloc0_n(column_count, sizeof(inc->column_details[0]));
394 column_idx = channel_idx = 0;
395 for (format_idx = 0; format_idx < format_count; format_idx++) {
396 format = formats[format_idx];
397 (void)split_column_format(format, &c, &f, &b);
399 detail = &inc->column_details[column_idx++];
400 detail->col_nr = column_idx;
401 detail->text_format = f;
402 if (detail->text_format) {
403 detail->channel_offset = channel_idx;
404 detail->channel_count = b;
407 sr_dbg("detail -> col %zu, fmt %s, ch off/cnt %zu/%zu",
408 detail->col_nr, col_format_text[detail->text_format],
409 detail->channel_offset, detail->channel_count);
412 inc->logic_channels = channel_idx;
418 static int make_column_details_single(struct context *inc,
419 size_t col_nr, size_t bit_count, enum single_col_format format)
421 struct column_details *details;
424 * Need at least as many columns to also include the one with
425 * the single-column input data.
427 inc->column_want_count = col_nr;
430 * Allocate the columns' processing details. Columns are counted
431 * from 1 (user's perspective), array items from 0 (programmer's
434 inc->column_details = g_malloc0_n(col_nr, sizeof(inc->column_details[0]));
435 details = &inc->column_details[col_nr - 1];
436 details->col_nr = col_nr;
439 * In single-column mode this single column will hold all bits
440 * of all logic channels, in the user specified number format.
442 details->text_format = format;
443 details->channel_offset = 0;
444 details->channel_count = bit_count;
449 static int make_column_details_multi(struct context *inc,
450 size_t first_col, size_t last_col)
452 struct column_details *details;
456 * Need at least as many columns to also include the one with
457 * the last channel's data.
459 inc->column_want_count = last_col;
462 * Allocate the columns' processing details. Columns are counted
463 * from 1, array items from 0.
464 * In multi-column mode each column will hold a single bit for
465 * the respective channel.
467 inc->column_details = g_malloc0_n(last_col, sizeof(inc->column_details[0]));
468 for (col_nr = first_col; col_nr <= last_col; col_nr++) {
469 details = &inc->column_details[col_nr - 1];
470 details->col_nr = col_nr;
471 details->text_format = FORMAT_BIN;
472 details->channel_offset = col_nr - first_col;
473 details->channel_count = 1;
480 static const struct column_details *lookup_column_details(struct context *inc, size_t nr)
482 if (!inc || !inc->column_details)
484 if (!nr || nr > inc->column_want_count)
486 return &inc->column_details[nr - 1];
490 * Primitive operations for text input: Strip comments off text lines.
491 * Split text lines into columns. Process input text for individual
495 static void strip_comment(char *buf, const GString *prefix)
502 if ((ptr = strstr(buf, prefix->str))) {
509 * @brief Splits a text line into a set of columns.
511 * @param[in] buf The input text line to split.
512 * @param[in] inc The input module's context.
514 * @returns An array of strings, representing the columns' text.
516 * This routine splits a text line on previously determined separators.
518 static char **split_line(char *buf, struct context *inc)
520 return g_strsplit(buf, inc->delimiter->str, 0);
524 * @brief Parse a multi-bit field into several logic channels.
526 * @param[in] column The input text, a run of bin/hex/oct digits.
527 * @param[in] inc The input module's context.
528 * @param[in] details The column processing details.
530 * @retval SR_OK Success.
531 * @retval SR_ERR Invalid input data (empty, or format error).
533 * This routine modifies the logic levels in the current sample set,
534 * based on the text input and a user provided format spec.
536 static int parse_logic(const char *column, struct context *inc,
537 const struct column_details *details)
539 size_t length, ch_rem, ch_idx, ch_inc;
543 const char *type_text;
547 * Prepare to read the digits from the text end towards the start.
548 * A digit corresponds to a variable number of channels (depending
549 * on the value's radix). Prepare the mapping of text digits to
550 * (a number of) logic channels.
552 length = strlen(column);
554 sr_err("Column %zu in line %zu is empty.", details->col_nr,
558 rdptr = &column[length];
559 ch_idx = details->channel_offset;
560 ch_rem = details->channel_count;
563 * Get another digit and derive up to four logic channels' state from
564 * it. Make sure to not process more bits than the column has channels
565 * associated with it.
567 while (rdptr > column && ch_rem) {
568 /* Check for valid digits according to the input radix. */
570 switch (details->text_format) {
572 valid = g_ascii_isxdigit(c) && c < '2';
576 valid = g_ascii_isxdigit(c) && c < '8';
580 valid = g_ascii_isxdigit(c);
588 type_text = col_format_text[details->text_format];
589 sr_err("Invalid text '%s' in %s type column %zu in line %zu.",
590 column, type_text, details->col_nr, inc->line_number);
593 /* Use the digit's bits for logic channels' data. */
594 bits = g_ascii_xdigit_value(c);
595 switch (details->text_format) {
599 set_logic_level(inc, ch_idx + 3, bits & (1 << 3));
605 set_logic_level(inc, ch_idx + 2, bits & (1 << 2));
609 set_logic_level(inc, ch_idx + 1, bits & (1 << 1));
614 set_logic_level(inc, ch_idx + 0, bits & (1 << 0));
617 /* ShouldNotHappen(TM), but silences compiler warning. */
623 * TODO Determine whether the availability of extra input data
624 * for unhandled logic channels is worth warning here. In this
625 * implementation users are in control, and can have the more
626 * significant bits ignored (which can be considered a feature
627 * and not really a limitation).
634 * @brief Parse routine which ignores the input text.
636 * This routine exists to unify dispatch code paths, mapping input file
637 * columns' data types to their respective parse routines.
639 static int parse_ignore(const char *column, struct context *inc,
640 const struct column_details *details)
648 typedef int (*col_parse_cb)(const char *column, struct context *inc,
649 const struct column_details *details);
651 static const col_parse_cb col_parse_funcs[] = {
652 [FORMAT_NONE] = parse_ignore,
653 [FORMAT_BIN] = parse_logic,
654 [FORMAT_OCT] = parse_logic,
655 [FORMAT_HEX] = parse_logic,
658 static int init(struct sr_input *in, GHashTable *options)
661 size_t single_column;
663 enum single_col_format format;
666 in->sdi = g_malloc0(sizeof(*in->sdi));
667 in->priv = inc = g_malloc0(sizeof(*inc));
669 single_column = g_variant_get_uint32(g_hash_table_lookup(options, "single-column"));
670 inc->multi_column_mode = single_column == 0;
672 inc->logic_channels = g_variant_get_uint32(g_hash_table_lookup(options, "numchannels"));
674 inc->delimiter = g_string_new(g_variant_get_string(
675 g_hash_table_lookup(options, "delimiter"), NULL));
676 if (!inc->delimiter->len) {
677 sr_err("Column delimiter cannot be empty.");
681 s = g_variant_get_string(g_hash_table_lookup(options, "format"), NULL);
682 if (g_ascii_strncasecmp(s, "bin", 3) == 0) {
684 } else if (g_ascii_strncasecmp(s, "hex", 3) == 0) {
686 } else if (g_ascii_strncasecmp(s, "oct", 3) == 0) {
689 sr_err("Invalid format: '%s'", s);
693 inc->comment = g_string_new(g_variant_get_string(
694 g_hash_table_lookup(options, "comment"), NULL));
695 if (g_string_equal(inc->comment, inc->delimiter)) {
697 * Using the same sequence as comment leader and column
698 * delimiter won't work. The user probably specified ';'
699 * as the column delimiter but did not adjust the comment
700 * leader. Try DWIM, drop comment strippin support here.
702 sr_warn("Comment leader and column delimiter conflict, disabling comment support.");
703 g_string_truncate(inc->comment, 0);
706 inc->samplerate = g_variant_get_uint64(g_hash_table_lookup(options, "samplerate"));
708 inc->first_column = g_variant_get_uint32(g_hash_table_lookup(options, "first-column"));
710 inc->use_header = g_variant_get_boolean(g_hash_table_lookup(options, "header"));
712 inc->start_line = g_variant_get_uint32(g_hash_table_lookup(options, "startline"));
713 if (inc->start_line < 1) {
714 sr_err("Invalid start line %zu.", inc->start_line);
719 * Derive the set of columns to inspect and their respective
720 * formats from simple input specs. Remain close to the previous
721 * set of option keywords and their meaning. Exclusively support
722 * a single column with multiple bits in it, or an adjacent set
723 * of colums with one bit each. The latter may not know the total
724 * column count here (when the user omitted the spec), and will
725 * derive it from the first text line of the input file.
727 s = g_variant_get_string(g_hash_table_lookup(options, "column-formats"), NULL);
729 ret = make_column_details_from_format(inc, s);
732 } else if (single_column && inc->logic_channels) {
733 sr_dbg("DIAG Got single column (%zu) and channels (%zu).",
734 single_column, inc->logic_channels);
735 sr_dbg("DIAG -> column %zu, %zu bits in %s format.",
736 single_column, inc->logic_channels,
737 col_format_text[format]);
738 ret = make_column_details_single(inc,
739 single_column, inc->logic_channels, format);
742 } else if (inc->multi_column_mode) {
743 sr_dbg("DIAG Got multi-column, first column %zu, count %zu.",
744 inc->first_column, inc->logic_channels);
745 if (inc->logic_channels) {
746 sr_dbg("DIAG -> columns %zu-%zu, 1 bit each.",
748 inc->first_column + inc->logic_channels - 1);
749 ret = make_column_details_multi(inc, inc->first_column,
750 inc->first_column + inc->logic_channels - 1);
754 sr_dbg("DIAG -> incomplete spec, have to update later.");
757 sr_err("Unknown or unsupported combination of option values.");
765 * Check the channel list for consistency across file re-import. See
766 * the VCD input module for more details and motivation.
769 static void keep_header_for_reread(const struct sr_input *in)
774 g_slist_free_full(inc->prev_sr_channels, sr_channel_free_cb);
775 inc->prev_sr_channels = in->sdi->channels;
776 in->sdi->channels = NULL;
779 static int check_header_in_reread(const struct sr_input *in)
788 if (!inc->prev_sr_channels)
791 if (sr_channel_lists_differ(inc->prev_sr_channels, in->sdi->channels)) {
792 sr_err("Channel list change not supported for file re-read.");
795 g_slist_free_full(in->sdi->channels, sr_channel_free_cb);
796 in->sdi->channels = inc->prev_sr_channels;
797 inc->prev_sr_channels = NULL;
802 static const char *delim_set = "\r\n";
804 static const char *get_line_termination(GString *buf)
809 if (g_strstr_len(buf->str, buf->len, "\r\n"))
811 else if (memchr(buf->str, '\n', buf->len))
813 else if (memchr(buf->str, '\r', buf->len))
819 static int initial_parse(const struct sr_input *in, GString *buf)
822 GString *channel_name;
823 size_t num_columns, ch_idx, ch_name_idx, col_idx, col_nr;
824 size_t line_number, line_idx;
826 char **lines, *line, **columns, *column;
827 const char *col_caption;
828 gboolean got_caption;
829 const struct column_details *detail;
836 if (inc->termination)
837 lines = g_strsplit(buf->str, inc->termination, 0);
839 lines = g_strsplit_set(buf->str, delim_set, 0);
840 for (line_idx = 0; (line = lines[line_idx]); line_idx++) {
842 if (inc->start_line > line_number) {
843 sr_spew("Line %zu skipped (before start).", line_number);
846 if (line[0] == '\0') {
847 sr_spew("Blank line %zu skipped.", line_number);
850 strip_comment(line, inc->comment);
851 if (line[0] == '\0') {
852 sr_spew("Comment-only line %zu skipped.", line_number);
856 /* Reached first proper line. */
860 /* Not enough data for a proper line yet. */
865 /* See how many columns the current line has. */
866 columns = split_line(line, inc);
868 sr_err("Error while parsing line %zu.", line_number);
872 num_columns = g_strv_length(columns);
874 sr_err("Error while parsing line %zu.", line_number);
878 sr_dbg("DIAG Got %zu columns in text line: %s.", num_columns, line);
880 /* Optionally update incomplete multi-column specs. */
881 if (inc->multi_column_mode && !inc->logic_channels) {
882 inc->logic_channels = num_columns - inc->first_column + 1;
883 sr_dbg("DIAG -> multi-column update: columns %zu-%zu, 1 bit each.",
885 inc->first_column + inc->logic_channels - 1);
886 ret = make_column_details_multi(inc, inc->first_column,
887 inc->first_column + inc->logic_channels - 1);
893 * Assume all lines have equal length (column count). Bail out
894 * early on suspicious or insufficient input data (check input
895 * which became available here against previous user specs or
896 * auto-determined properties, regardless of layout variant).
898 if (num_columns < inc->column_want_count) {
899 sr_err("Insufficient input text width for desired data amount, got %zu but want %zu columns.",
900 num_columns, inc->column_want_count);
906 * Determine channel names. Optionally use text from a header
907 * line (when requested by the user, and only works in multi
908 * column mode). In the absence of header text, or in single
909 * column mode, channels are assigned rather generic names.
911 * Manipulation of the column's caption is acceptable here, the
912 * header line will never get processed another time.
914 channel_name = g_string_sized_new(64);
915 for (col_idx = 0; col_idx < inc->column_want_count; col_idx++) {
917 col_nr = col_idx + 1;
918 detail = lookup_column_details(inc, col_nr);
919 if (detail->text_format == FORMAT_NONE)
921 column = columns[col_idx];
922 col_caption = sr_scpi_unquote_string(column);
923 got_caption = inc->use_header && *col_caption;
924 sr_dbg("DIAG col %zu, ch count %zu, text %s.",
925 col_nr, detail->channel_count, col_caption);
926 for (ch_idx = 0; ch_idx < detail->channel_count; ch_idx++) {
927 ch_name_idx = detail->channel_offset + ch_idx;
928 if (got_caption && detail->channel_count == 1)
929 g_string_assign(channel_name, col_caption);
930 else if (got_caption)
931 g_string_printf(channel_name, "%s[%zu]",
932 col_caption, ch_idx);
934 g_string_printf(channel_name, "%zu", ch_name_idx);
935 sr_dbg("DIAG ch idx %zu, name %s.", ch_name_idx, channel_name->str);
936 sr_channel_new(in->sdi, ch_name_idx, SR_CHANNEL_LOGIC, TRUE,
940 g_string_free(channel_name, TRUE);
941 if (!check_header_in_reread(in)) {
947 * Calculate the minimum buffer size to store the set of samples
948 * of all channels (unit size). Determine a larger buffer size
949 * for datafeed submission that is a multiple of the unit size.
950 * Allocate the larger buffer, the "sample buffer" will point
951 * to a location within that large buffer later.
953 inc->sample_unit_size = (inc->logic_channels + 7) / 8;
954 inc->datafeed_buf_size = CHUNK_SIZE;
955 inc->datafeed_buf_size *= inc->sample_unit_size;
956 inc->datafeed_buffer = g_malloc(inc->datafeed_buf_size);
957 inc->datafeed_buf_fill = 0;
968 * Gets called from initial_receive(), which runs until the end-of-line
969 * encoding of the input stream could get determined. Assumes that this
970 * routine receives enough buffered initial input data to either see the
971 * BOM when there is one, or that no BOM will follow when a text line
972 * termination sequence was seen. Silently drops the UTF-8 BOM sequence
973 * from the input buffer if one was seen. Does not care to protect
974 * against multiple execution or dropping the BOM multiple times --
975 * there should be at most one in the input stream.
977 static void initial_bom_check(const struct sr_input *in)
979 static const char *utf8_bom = "\xef\xbb\xbf";
981 if (in->buf->len < strlen(utf8_bom))
983 if (strncmp(in->buf->str, utf8_bom, strlen(utf8_bom)) != 0)
985 g_string_erase(in->buf, 0, strlen(utf8_bom));
988 static int initial_receive(const struct sr_input *in)
994 const char *termination;
996 initial_bom_check(in);
1000 termination = get_line_termination(in->buf);
1002 /* Don't have a full line yet. */
1005 p = g_strrstr_len(in->buf->str, in->buf->len, termination);
1007 /* Don't have a full line yet. */
1009 len = p - in->buf->str - 1;
1010 new_buf = g_string_new_len(in->buf->str, len);
1011 g_string_append_c(new_buf, '\0');
1013 inc->termination = g_strdup(termination);
1015 if (in->buf->str[0] != '\0')
1016 ret = initial_parse(in, new_buf);
1020 g_string_free(new_buf, TRUE);
1025 static int process_buffer(struct sr_input *in, gboolean is_eof)
1027 struct context *inc;
1029 size_t line_idx, col_idx, col_nr;
1030 const struct column_details *details;
1031 col_parse_cb parse_func;
1033 char *p, **lines, *line, **columns, *column;
1036 if (!inc->started) {
1037 std_session_send_df_header(in->sdi);
1038 inc->started = TRUE;
1042 * Consider empty input non-fatal. Keep accumulating input until
1043 * at least one full text line has become available. Grab the
1044 * maximum amount of accumulated data that consists of full text
1045 * lines, and process what has been received so far, leaving not
1046 * yet complete lines for the next invocation.
1048 * Enforce that all previously buffered data gets processed in
1049 * the "EOF" condition. Do not insist in the presence of the
1050 * termination sequence for the last line (may often be missing
1051 * on Windows). A present termination sequence will just result
1052 * in the "execution of an empty line", and does not harm.
1057 p = in->buf->str + in->buf->len;
1059 p = g_strrstr_len(in->buf->str, in->buf->len, inc->termination);
1063 p += strlen(inc->termination);
1065 g_strstrip(in->buf->str);
1068 lines = g_strsplit(in->buf->str, inc->termination, 0);
1069 for (line_idx = 0; (line = lines[line_idx]); line_idx++) {
1071 if (inc->line_number < inc->start_line) {
1072 sr_spew("Line %zu skipped (before start).", inc->line_number);
1075 if (line[0] == '\0') {
1076 sr_spew("Blank line %zu skipped.", inc->line_number);
1080 /* Remove trailing comment. */
1081 strip_comment(line, inc->comment);
1082 if (line[0] == '\0') {
1083 sr_spew("Comment-only line %zu skipped.", inc->line_number);
1087 /* Skip the header line, its content was used as the channel names. */
1088 if (inc->use_header && !inc->header_seen) {
1089 sr_spew("Header line %zu skipped.", inc->line_number);
1090 inc->header_seen = TRUE;
1094 /* Split the line into columns, check for minimum length. */
1095 columns = split_line(line, inc);
1097 sr_err("Error while parsing line %zu.", inc->line_number);
1101 num_columns = g_strv_length(columns);
1102 if (num_columns < inc->column_want_count) {
1103 sr_err("Insufficient column count %zu in line %zu.",
1104 num_columns, inc->line_number);
1105 g_strfreev(columns);
1110 /* Have the columns of the current text line processed. */
1111 clear_logic_samples(inc);
1112 for (col_idx = 0; col_idx < inc->column_want_count; col_idx++) {
1113 column = columns[col_idx];
1114 col_nr = col_idx + 1;
1115 details = lookup_column_details(inc, col_nr);
1116 if (!details || !details->text_format)
1118 parse_func = col_parse_funcs[details->text_format];
1121 ret = parse_func(column, inc, details);
1123 g_strfreev(columns);
1129 /* Send sample data to the session bus (buffered). */
1130 ret = queue_logic_samples(in);
1132 sr_err("Sending samples failed.");
1133 g_strfreev(columns);
1138 g_strfreev(columns);
1141 g_string_erase(in->buf, 0, p - in->buf->str);
1146 static int receive(struct sr_input *in, GString *buf)
1148 struct context *inc;
1151 g_string_append_len(in->buf, buf->str, buf->len);
1154 if (!inc->termination) {
1155 ret = initial_receive(in);
1156 if (ret == SR_ERR_NA)
1157 /* Not enough data yet. */
1159 else if (ret != SR_OK)
1162 /* sdi is ready, notify frontend. */
1163 in->sdi_ready = TRUE;
1167 ret = process_buffer(in, FALSE);
1172 static int end(struct sr_input *in)
1174 struct context *inc;
1178 ret = process_buffer(in, TRUE);
1184 ret = flush_logic_samples(in);
1190 std_session_send_df_end(in->sdi);
1195 static void cleanup(struct sr_input *in)
1197 struct context *inc;
1199 keep_header_for_reread(in);
1203 g_free(inc->termination);
1204 inc->termination = NULL;
1205 g_free(inc->datafeed_buffer);
1206 inc->datafeed_buffer = NULL;
1209 static int reset(struct sr_input *in)
1211 struct context *inc = in->priv;
1214 inc->started = FALSE;
1215 g_string_truncate(in->buf, 0);
1234 static struct sr_option options[] = {
1235 [OPT_COL_FMTS] = { "column-formats", "Column format specs", "Specifies text columns data types: comma separated list of [<cols>]<fmt>[<bits>]", NULL, NULL },
1236 [OPT_SINGLE_COL] = { "single-column", "Single column", "Enable single-column mode, using the specified column (>= 1); 0: multi-col. mode", NULL, NULL },
1237 [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 },
1238 [OPT_DELIM] = { "delimiter", "Column delimiter", "The column delimiter (>= 1 characters)", NULL, NULL },
1239 [OPT_FORMAT] = { "format", "Data format (single-col. mode)", "The numeric format of the data (single-col. mode): bin, hex, oct", NULL, NULL },
1240 [OPT_COMMENT] = { "comment", "Comment character(s)", "The comment prefix character(s)", NULL, NULL },
1241 [OPT_RATE] = { "samplerate", "Samplerate (Hz)", "The sample rate (used during capture) in Hz", NULL, NULL },
1242 [OPT_FIRST_COL] = { "first-column", "First column", "The column number of the first channel (multi-col. mode)", NULL, NULL },
1243 [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 },
1244 [OPT_START] = { "startline", "Start line", "The line number at which to start processing samples (>= 1)", NULL, NULL },
1245 [OPT_MAX] = ALL_ZERO,
1248 static const struct sr_option *get_options(void)
1252 if (!options[0].def) {
1253 options[OPT_COL_FMTS].def = g_variant_ref_sink(g_variant_new_string("0l"));
1254 options[OPT_SINGLE_COL].def = g_variant_ref_sink(g_variant_new_uint32(0));
1255 options[OPT_NUM_LOGIC].def = g_variant_ref_sink(g_variant_new_uint32(0));
1256 options[OPT_DELIM].def = g_variant_ref_sink(g_variant_new_string(","));
1257 options[OPT_FORMAT].def = g_variant_ref_sink(g_variant_new_string("bin"));
1259 l = g_slist_append(l, g_variant_ref_sink(g_variant_new_string("bin")));
1260 l = g_slist_append(l, g_variant_ref_sink(g_variant_new_string("hex")));
1261 l = g_slist_append(l, g_variant_ref_sink(g_variant_new_string("oct")));
1262 options[OPT_FORMAT].values = l;
1263 options[OPT_COMMENT].def = g_variant_ref_sink(g_variant_new_string(";"));
1264 options[OPT_RATE].def = g_variant_ref_sink(g_variant_new_uint64(0));
1265 options[OPT_FIRST_COL].def = g_variant_ref_sink(g_variant_new_uint32(1));
1266 options[OPT_HEADER].def = g_variant_ref_sink(g_variant_new_boolean(FALSE));
1267 options[OPT_START].def = g_variant_ref_sink(g_variant_new_uint32(1));
1273 SR_PRIV struct sr_input_module input_csv = {
1276 .desc = "Comma-separated values",
1277 .exts = (const char*[]){"csv", NULL},
1278 .options = get_options,