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 static const char col_format_char[] = {
131 struct column_details {
133 enum single_col_format text_format;
134 size_t channel_offset;
135 size_t channel_count;
141 /* Current selected samplerate. */
143 gboolean samplerate_sent;
145 /* Number of logic channels. */
146 size_t logic_channels;
148 /* Column delimiter (actually separator), comment leader, EOL sequence. */
153 /* Format specs for input columns, and processing state. */
154 size_t column_seen_count;
155 const char *column_formats;
156 size_t column_want_count;
157 struct column_details *column_details;
159 /* Line number to start processing. */
163 * Determines if the first line should be treated as header and used for
164 * channel names in multi column mode.
167 gboolean header_seen;
169 size_t sample_unit_size; /**!< Byte count for a single sample. */
170 uint8_t *sample_buffer; /**!< Buffer for a single sample. */
172 uint8_t *datafeed_buffer; /**!< Queue for datafeed submission. */
173 size_t datafeed_buf_size;
174 size_t datafeed_buf_fill;
176 /* Current line number. */
179 /* List of previously created sigrok channels. */
180 GSList *prev_sr_channels;
184 * Primitive operations to handle sample sets:
185 * - Keep a buffer for datafeed submission, capable of holding many
186 * samples (reduces call overhead, improves throughput).
187 * - Have a "current sample set" pointer reference one position in that
188 * large samples buffer.
189 * - Clear the current sample set before text line inspection, then set
190 * the bits which are found active in the current line of text input.
191 * Phrase the API such that call sites can be kept simple. Advance to
192 * the next sample set between lines, flush the larger buffer as needed
193 * (when it is full, or upon EOF).
196 static void clear_logic_samples(struct context *inc)
198 inc->sample_buffer = &inc->datafeed_buffer[inc->datafeed_buf_fill];
199 memset(inc->sample_buffer, 0, inc->sample_unit_size);
202 static void set_logic_level(struct context *inc, size_t ch_idx, int on)
204 size_t byte_idx, bit_idx;
207 if (ch_idx >= inc->logic_channels)
212 byte_idx = ch_idx / 8;
213 bit_idx = ch_idx % 8;
214 bit_mask = 1 << bit_idx;
215 inc->sample_buffer[byte_idx] |= bit_mask;
218 static int flush_logic_samples(const struct sr_input *in)
221 struct sr_datafeed_packet packet;
222 struct sr_datafeed_meta meta;
223 struct sr_config *src;
225 struct sr_datafeed_logic logic;
229 if (!inc->datafeed_buf_fill)
232 if (inc->samplerate && !inc->samplerate_sent) {
233 packet.type = SR_DF_META;
234 packet.payload = &meta;
235 samplerate = inc->samplerate;
236 src = sr_config_new(SR_CONF_SAMPLERATE, g_variant_new_uint64(samplerate));
237 meta.config = g_slist_append(NULL, src);
238 sr_session_send(in->sdi, &packet);
239 g_slist_free(meta.config);
241 inc->samplerate_sent = TRUE;
244 memset(&packet, 0, sizeof(packet));
245 memset(&logic, 0, sizeof(logic));
246 packet.type = SR_DF_LOGIC;
247 packet.payload = &logic;
248 logic.unitsize = inc->sample_unit_size;
249 logic.length = inc->datafeed_buf_fill;
250 logic.data = inc->datafeed_buffer;
252 rc = sr_session_send(in->sdi, &packet);
256 inc->datafeed_buf_fill = 0;
260 static int queue_logic_samples(const struct sr_input *in)
266 if (!inc->logic_channels)
269 inc->datafeed_buf_fill += inc->sample_unit_size;
270 if (inc->datafeed_buf_fill == inc->datafeed_buf_size) {
271 rc = flush_logic_samples(in);
278 /* Helpers for "column processing". */
280 static int split_column_format(const char *spec,
281 size_t *column_count, enum single_col_format *format, size_t *bit_count)
284 char *endp, format_char;
285 enum single_col_format format_code;
290 /* Get the (optional, decimal, default 1) column count. Accept '*'. */
294 endp = (char *)&spec[1];
296 count = strtoul(spec, &endp, 10);
303 *column_count = count;
306 /* Get the (mandatory, single letter) type spec (-/xob/l). */
307 format_char = *spec++;
308 switch (format_char) {
309 case '-': /* Might conflict with number-parsing. */
312 format_code = FORMAT_NONE;
315 format_code = FORMAT_HEX;
318 format_code = FORMAT_OCT;
322 format_code = FORMAT_BIN;
324 default: /* includes NUL */
328 *format = format_code;
330 /* Get the (optional, decimal, default 1) bit count. */
332 count = strtoul(spec, &endp, 10);
337 if (format_char == '-')
339 if (format_char == 'l')
345 /* Input spec must have been exhausted. */
352 static int make_column_details_from_format(struct context *inc,
353 const char *column_format)
355 char **formats, *format;
356 size_t format_count, column_count, bit_count;
357 size_t auto_column_count;
358 size_t format_idx, c, b, column_idx, channel_idx;
359 enum single_col_format f;
360 struct column_details *detail;
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 auto_column_count = 0;
377 for (format_idx = 0; format_idx < format_count; format_idx++) {
378 format = formats[format_idx];
379 ret = split_column_format(format, &c, &f, &b);
380 sr_dbg("fmt %s -> %zu cols, %s fmt, %zu bits, rc %d", format, c, col_format_text[f], b, ret);
382 sr_err("Cannot parse columns format %s (field split, %s).", column_format, format);
387 /* User requested "auto-count", must be last format. */
388 if (formats[format_idx + 1]) {
389 sr_err("Auto column count must be last format field.");
393 auto_column_count = inc->column_seen_count - column_count;
394 c = auto_column_count;
399 sr_dbg("Column format %s -> %zu columns, %zu logic channels.",
400 column_format, column_count, bit_count);
402 /* Allocate and fill in "column processing" details. */
403 inc->column_want_count = column_count;
404 inc->column_details = g_malloc0_n(column_count, sizeof(inc->column_details[0]));
405 column_idx = channel_idx = 0;
406 for (format_idx = 0; format_idx < format_count; format_idx++) {
407 format = formats[format_idx];
408 (void)split_column_format(format, &c, &f, &b);
410 c = auto_column_count;
412 detail = &inc->column_details[column_idx++];
413 detail->col_nr = column_idx;
414 detail->text_format = f;
415 if (detail->text_format) {
416 detail->channel_offset = channel_idx;
417 detail->channel_count = b;
420 sr_dbg("detail -> col %zu, fmt %s, ch off/cnt %zu/%zu",
421 detail->col_nr, col_format_text[detail->text_format],
422 detail->channel_offset, detail->channel_count);
425 inc->logic_channels = channel_idx;
431 static const struct column_details *lookup_column_details(struct context *inc, size_t nr)
433 if (!inc || !inc->column_details)
435 if (!nr || nr > inc->column_want_count)
437 return &inc->column_details[nr - 1];
441 * Primitive operations for text input: Strip comments off text lines.
442 * Split text lines into columns. Process input text for individual
446 static void strip_comment(char *buf, const GString *prefix)
453 if ((ptr = strstr(buf, prefix->str))) {
460 * @brief Splits a text line into a set of columns.
462 * @param[in] buf The input text line to split.
463 * @param[in] inc The input module's context.
465 * @returns An array of strings, representing the columns' text.
467 * This routine splits a text line on previously determined separators.
469 static char **split_line(char *buf, struct context *inc)
471 return g_strsplit(buf, inc->delimiter->str, 0);
475 * @brief Parse a multi-bit field into several logic channels.
477 * @param[in] column The input text, a run of bin/hex/oct digits.
478 * @param[in] inc The input module's context.
479 * @param[in] details The column processing details.
481 * @retval SR_OK Success.
482 * @retval SR_ERR Invalid input data (empty, or format error).
484 * This routine modifies the logic levels in the current sample set,
485 * based on the text input and a user provided format spec.
487 static int parse_logic(const char *column, struct context *inc,
488 const struct column_details *details)
490 size_t length, ch_rem, ch_idx, ch_inc;
494 const char *type_text;
498 * Prepare to read the digits from the text end towards the start.
499 * A digit corresponds to a variable number of channels (depending
500 * on the value's radix). Prepare the mapping of text digits to
501 * (a number of) logic channels.
503 length = strlen(column);
505 sr_err("Column %zu in line %zu is empty.", details->col_nr,
509 rdptr = &column[length];
510 ch_idx = details->channel_offset;
511 ch_rem = details->channel_count;
514 * Get another digit and derive up to four logic channels' state from
515 * it. Make sure to not process more bits than the column has channels
516 * associated with it.
518 while (rdptr > column && ch_rem) {
519 /* Check for valid digits according to the input radix. */
521 switch (details->text_format) {
523 valid = g_ascii_isxdigit(c) && c < '2';
527 valid = g_ascii_isxdigit(c) && c < '8';
531 valid = g_ascii_isxdigit(c);
539 type_text = col_format_text[details->text_format];
540 sr_err("Invalid text '%s' in %s type column %zu in line %zu.",
541 column, type_text, details->col_nr, inc->line_number);
544 /* Use the digit's bits for logic channels' data. */
545 bits = g_ascii_xdigit_value(c);
546 switch (details->text_format) {
550 set_logic_level(inc, ch_idx + 3, bits & (1 << 3));
556 set_logic_level(inc, ch_idx + 2, bits & (1 << 2));
560 set_logic_level(inc, ch_idx + 1, bits & (1 << 1));
565 set_logic_level(inc, ch_idx + 0, bits & (1 << 0));
568 /* ShouldNotHappen(TM), but silences compiler warning. */
574 * TODO Determine whether the availability of extra input data
575 * for unhandled logic channels is worth warning here. In this
576 * implementation users are in control, and can have the more
577 * significant bits ignored (which can be considered a feature
578 * and not really a limitation).
585 * @brief Parse routine which ignores the input text.
587 * This routine exists to unify dispatch code paths, mapping input file
588 * columns' data types to their respective parse routines.
590 static int parse_ignore(const char *column, struct context *inc,
591 const struct column_details *details)
599 typedef int (*col_parse_cb)(const char *column, struct context *inc,
600 const struct column_details *details);
602 static const col_parse_cb col_parse_funcs[] = {
603 [FORMAT_NONE] = parse_ignore,
604 [FORMAT_BIN] = parse_logic,
605 [FORMAT_OCT] = parse_logic,
606 [FORMAT_HEX] = parse_logic,
609 static int init(struct sr_input *in, GHashTable *options)
612 size_t single_column, first_column, logic_channels;
614 enum single_col_format format;
617 in->sdi = g_malloc0(sizeof(*in->sdi));
618 in->priv = inc = g_malloc0(sizeof(*inc));
620 single_column = g_variant_get_uint32(g_hash_table_lookup(options, "single-column"));
622 logic_channels = g_variant_get_uint32(g_hash_table_lookup(options, "numchannels"));
624 inc->delimiter = g_string_new(g_variant_get_string(
625 g_hash_table_lookup(options, "delimiter"), NULL));
626 if (!inc->delimiter->len) {
627 sr_err("Column delimiter cannot be empty.");
631 s = g_variant_get_string(g_hash_table_lookup(options, "format"), NULL);
632 if (g_ascii_strncasecmp(s, "bin", 3) == 0) {
634 } else if (g_ascii_strncasecmp(s, "hex", 3) == 0) {
636 } else if (g_ascii_strncasecmp(s, "oct", 3) == 0) {
639 sr_err("Invalid format: '%s'", s);
643 inc->comment = g_string_new(g_variant_get_string(
644 g_hash_table_lookup(options, "comment"), NULL));
645 if (g_string_equal(inc->comment, inc->delimiter)) {
647 * Using the same sequence as comment leader and column
648 * delimiter won't work. The user probably specified ';'
649 * as the column delimiter but did not adjust the comment
650 * leader. Try DWIM, drop comment strippin support here.
652 sr_warn("Comment leader and column delimiter conflict, disabling comment support.");
653 g_string_truncate(inc->comment, 0);
656 inc->samplerate = g_variant_get_uint64(g_hash_table_lookup(options, "samplerate"));
658 first_column = g_variant_get_uint32(g_hash_table_lookup(options, "first-column"));
660 inc->use_header = g_variant_get_boolean(g_hash_table_lookup(options, "header"));
662 inc->start_line = g_variant_get_uint32(g_hash_table_lookup(options, "startline"));
663 if (inc->start_line < 1) {
664 sr_err("Invalid start line %zu.", inc->start_line);
669 * Scan flexible, to get prefered format specs which describe
670 * the input file's data formats. As well as some simple specs
671 * for backwards compatibility and user convenience.
673 * This logic ends up with a copy of the format string, either
674 * user provided or internally derived. Actual creation of the
675 * column processing details gets deferred until the first line
676 * of input data was seen. To support automatic determination of
677 * e.g. channel counts from column counts.
679 s = g_variant_get_string(g_hash_table_lookup(options, "column-formats"), NULL);
681 inc->column_formats = g_strdup(s);
682 sr_dbg("User specified column-formats: %s.", s);
683 } else if (single_column && logic_channels) {
684 format_char = col_format_char[format];
685 if (single_column == 1) {
686 inc->column_formats = g_strdup_printf("%c%zu",
687 format_char, logic_channels);
689 inc->column_formats = g_strdup_printf("%zu-,%c%zu",
691 format_char, logic_channels);
693 sr_dbg("Backwards compat single-column, col %zu, fmt %s, bits %zu -> %s.",
694 single_column, col_format_text[format], logic_channels,
695 inc->column_formats);
696 } else if (!single_column) {
697 if (first_column > 1) {
698 inc->column_formats = g_strdup_printf("%zu-,%zul",
699 first_column - 1, logic_channels);
701 inc->column_formats = g_strdup_printf("%zul",
704 sr_dbg("Backwards compat multi-column, col %zu, chans %zu -> %s.",
705 first_column, logic_channels,
706 inc->column_formats);
708 sr_warn("Unknown or unsupported format spec, assuming trivial multi-column.");
709 inc->column_formats = g_strdup("*l");
716 * Check the channel list for consistency across file re-import. See
717 * the VCD input module for more details and motivation.
720 static void keep_header_for_reread(const struct sr_input *in)
725 g_slist_free_full(inc->prev_sr_channels, sr_channel_free_cb);
726 inc->prev_sr_channels = in->sdi->channels;
727 in->sdi->channels = NULL;
730 static int check_header_in_reread(const struct sr_input *in)
739 if (!inc->prev_sr_channels)
742 if (sr_channel_lists_differ(inc->prev_sr_channels, in->sdi->channels)) {
743 sr_err("Channel list change not supported for file re-read.");
746 g_slist_free_full(in->sdi->channels, sr_channel_free_cb);
747 in->sdi->channels = inc->prev_sr_channels;
748 inc->prev_sr_channels = NULL;
753 static const char *delim_set = "\r\n";
755 static const char *get_line_termination(GString *buf)
760 if (g_strstr_len(buf->str, buf->len, "\r\n"))
762 else if (memchr(buf->str, '\n', buf->len))
764 else if (memchr(buf->str, '\r', buf->len))
770 static int initial_parse(const struct sr_input *in, GString *buf)
773 GString *channel_name;
774 size_t num_columns, ch_idx, ch_name_idx, col_idx, col_nr;
775 size_t line_number, line_idx;
777 char **lines, *line, **columns, *column;
778 const char *col_caption;
779 gboolean got_caption;
780 const struct column_details *detail;
787 if (inc->termination)
788 lines = g_strsplit(buf->str, inc->termination, 0);
790 lines = g_strsplit_set(buf->str, delim_set, 0);
791 for (line_idx = 0; (line = lines[line_idx]); line_idx++) {
793 if (inc->start_line > line_number) {
794 sr_spew("Line %zu skipped (before start).", line_number);
797 if (line[0] == '\0') {
798 sr_spew("Blank line %zu skipped.", line_number);
801 strip_comment(line, inc->comment);
802 if (line[0] == '\0') {
803 sr_spew("Comment-only line %zu skipped.", line_number);
807 /* Reached first proper line. */
811 /* Not enough data for a proper line yet. */
816 /* See how many columns the current line has. */
817 columns = split_line(line, inc);
819 sr_err("Error while parsing line %zu.", line_number);
823 num_columns = g_strv_length(columns);
825 sr_err("Error while parsing line %zu.", line_number);
829 sr_dbg("DIAG Got %zu columns in text line: %s.", num_columns, line);
832 * Track the observed number of columns in the input file. Do
833 * process the previously gathered columns format spec now that
834 * automatic channel count can be dealt with.
836 inc->column_seen_count = num_columns;
837 ret = make_column_details_from_format(inc, inc->column_formats);
839 sr_err("Cannot parse columns format using line %zu.", line_number);
844 * Assume all lines have equal length (column count). Bail out
845 * early on suspicious or insufficient input data (check input
846 * which became available here against previous user specs or
847 * auto-determined properties, regardless of layout variant).
849 if (num_columns < inc->column_want_count) {
850 sr_err("Insufficient input text width for desired data amount, got %zu but want %zu columns.",
851 num_columns, inc->column_want_count);
857 * Determine channel names. Optionally use text from a header
858 * line (when requested by the user, and only works in multi
859 * column mode). In the absence of header text, or in single
860 * column mode, channels are assigned rather generic names.
862 * Manipulation of the column's caption is acceptable here, the
863 * header line will never get processed another time.
865 channel_name = g_string_sized_new(64);
866 for (col_idx = 0; col_idx < inc->column_want_count; col_idx++) {
868 col_nr = col_idx + 1;
869 detail = lookup_column_details(inc, col_nr);
870 if (detail->text_format == FORMAT_NONE)
872 column = columns[col_idx];
873 col_caption = sr_scpi_unquote_string(column);
874 got_caption = inc->use_header && *col_caption;
875 sr_dbg("DIAG col %zu, ch count %zu, text %s.",
876 col_nr, detail->channel_count, col_caption);
877 for (ch_idx = 0; ch_idx < detail->channel_count; ch_idx++) {
878 ch_name_idx = detail->channel_offset + ch_idx;
879 if (got_caption && detail->channel_count == 1)
880 g_string_assign(channel_name, col_caption);
881 else if (got_caption)
882 g_string_printf(channel_name, "%s[%zu]",
883 col_caption, ch_idx);
885 g_string_printf(channel_name, "%zu", ch_name_idx);
886 sr_dbg("DIAG ch idx %zu, name %s.", ch_name_idx, channel_name->str);
887 sr_channel_new(in->sdi, ch_name_idx, SR_CHANNEL_LOGIC, TRUE,
891 g_string_free(channel_name, TRUE);
892 if (!check_header_in_reread(in)) {
898 * Calculate the minimum buffer size to store the set of samples
899 * of all channels (unit size). Determine a larger buffer size
900 * for datafeed submission that is a multiple of the unit size.
901 * Allocate the larger buffer, the "sample buffer" will point
902 * to a location within that large buffer later.
904 inc->sample_unit_size = (inc->logic_channels + 7) / 8;
905 inc->datafeed_buf_size = CHUNK_SIZE;
906 inc->datafeed_buf_size *= inc->sample_unit_size;
907 inc->datafeed_buffer = g_malloc(inc->datafeed_buf_size);
908 inc->datafeed_buf_fill = 0;
919 * Gets called from initial_receive(), which runs until the end-of-line
920 * encoding of the input stream could get determined. Assumes that this
921 * routine receives enough buffered initial input data to either see the
922 * BOM when there is one, or that no BOM will follow when a text line
923 * termination sequence was seen. Silently drops the UTF-8 BOM sequence
924 * from the input buffer if one was seen. Does not care to protect
925 * against multiple execution or dropping the BOM multiple times --
926 * there should be at most one in the input stream.
928 static void initial_bom_check(const struct sr_input *in)
930 static const char *utf8_bom = "\xef\xbb\xbf";
932 if (in->buf->len < strlen(utf8_bom))
934 if (strncmp(in->buf->str, utf8_bom, strlen(utf8_bom)) != 0)
936 g_string_erase(in->buf, 0, strlen(utf8_bom));
939 static int initial_receive(const struct sr_input *in)
945 const char *termination;
947 initial_bom_check(in);
951 termination = get_line_termination(in->buf);
953 /* Don't have a full line yet. */
956 p = g_strrstr_len(in->buf->str, in->buf->len, termination);
958 /* Don't have a full line yet. */
960 len = p - in->buf->str - 1;
961 new_buf = g_string_new_len(in->buf->str, len);
962 g_string_append_c(new_buf, '\0');
964 inc->termination = g_strdup(termination);
966 if (in->buf->str[0] != '\0')
967 ret = initial_parse(in, new_buf);
971 g_string_free(new_buf, TRUE);
976 static int process_buffer(struct sr_input *in, gboolean is_eof)
980 size_t line_idx, col_idx, col_nr;
981 const struct column_details *details;
982 col_parse_cb parse_func;
984 char *p, **lines, *line, **columns, *column;
988 std_session_send_df_header(in->sdi);
993 * Consider empty input non-fatal. Keep accumulating input until
994 * at least one full text line has become available. Grab the
995 * maximum amount of accumulated data that consists of full text
996 * lines, and process what has been received so far, leaving not
997 * yet complete lines for the next invocation.
999 * Enforce that all previously buffered data gets processed in
1000 * the "EOF" condition. Do not insist in the presence of the
1001 * termination sequence for the last line (may often be missing
1002 * on Windows). A present termination sequence will just result
1003 * in the "execution of an empty line", and does not harm.
1008 p = in->buf->str + in->buf->len;
1010 p = g_strrstr_len(in->buf->str, in->buf->len, inc->termination);
1014 p += strlen(inc->termination);
1016 g_strstrip(in->buf->str);
1019 lines = g_strsplit(in->buf->str, inc->termination, 0);
1020 for (line_idx = 0; (line = lines[line_idx]); line_idx++) {
1022 if (inc->line_number < inc->start_line) {
1023 sr_spew("Line %zu skipped (before start).", inc->line_number);
1026 if (line[0] == '\0') {
1027 sr_spew("Blank line %zu skipped.", inc->line_number);
1031 /* Remove trailing comment. */
1032 strip_comment(line, inc->comment);
1033 if (line[0] == '\0') {
1034 sr_spew("Comment-only line %zu skipped.", inc->line_number);
1038 /* Skip the header line, its content was used as the channel names. */
1039 if (inc->use_header && !inc->header_seen) {
1040 sr_spew("Header line %zu skipped.", inc->line_number);
1041 inc->header_seen = TRUE;
1045 /* Split the line into columns, check for minimum length. */
1046 columns = split_line(line, inc);
1048 sr_err("Error while parsing line %zu.", inc->line_number);
1052 num_columns = g_strv_length(columns);
1053 if (num_columns < inc->column_want_count) {
1054 sr_err("Insufficient column count %zu in line %zu.",
1055 num_columns, inc->line_number);
1056 g_strfreev(columns);
1061 /* Have the columns of the current text line processed. */
1062 clear_logic_samples(inc);
1063 for (col_idx = 0; col_idx < inc->column_want_count; col_idx++) {
1064 column = columns[col_idx];
1065 col_nr = col_idx + 1;
1066 details = lookup_column_details(inc, col_nr);
1067 if (!details || !details->text_format)
1069 parse_func = col_parse_funcs[details->text_format];
1072 ret = parse_func(column, inc, details);
1074 g_strfreev(columns);
1080 /* Send sample data to the session bus (buffered). */
1081 ret = queue_logic_samples(in);
1083 sr_err("Sending samples failed.");
1084 g_strfreev(columns);
1089 g_strfreev(columns);
1092 g_string_erase(in->buf, 0, p - in->buf->str);
1097 static int receive(struct sr_input *in, GString *buf)
1099 struct context *inc;
1102 g_string_append_len(in->buf, buf->str, buf->len);
1105 if (!inc->column_seen_count) {
1106 ret = initial_receive(in);
1107 if (ret == SR_ERR_NA)
1108 /* Not enough data yet. */
1110 else if (ret != SR_OK)
1113 /* sdi is ready, notify frontend. */
1114 in->sdi_ready = TRUE;
1118 ret = process_buffer(in, FALSE);
1123 static int end(struct sr_input *in)
1125 struct context *inc;
1129 ret = process_buffer(in, TRUE);
1135 ret = flush_logic_samples(in);
1141 std_session_send_df_end(in->sdi);
1146 static void cleanup(struct sr_input *in)
1148 struct context *inc;
1150 keep_header_for_reread(in);
1154 g_free(inc->termination);
1155 inc->termination = NULL;
1156 g_free(inc->datafeed_buffer);
1157 inc->datafeed_buffer = NULL;
1160 static int reset(struct sr_input *in)
1162 struct context *inc = in->priv;
1165 inc->started = FALSE;
1166 g_string_truncate(in->buf, 0);
1185 static struct sr_option options[] = {
1186 [OPT_COL_FMTS] = { "column-formats", "Column format specs", "Specifies text columns data types: comma separated list of [<cols>]<fmt>[<bits>]", NULL, NULL },
1187 [OPT_SINGLE_COL] = { "single-column", "Single column", "Enable single-column mode, using the specified column (>= 1); 0: multi-col. mode", NULL, NULL },
1188 [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 },
1189 [OPT_DELIM] = { "delimiter", "Column delimiter", "The column delimiter (>= 1 characters)", NULL, NULL },
1190 [OPT_FORMAT] = { "format", "Data format (single-col. mode)", "The numeric format of the data (single-col. mode): bin, hex, oct", NULL, NULL },
1191 [OPT_COMMENT] = { "comment", "Comment character(s)", "The comment prefix character(s)", NULL, NULL },
1192 [OPT_RATE] = { "samplerate", "Samplerate (Hz)", "The sample rate (used during capture) in Hz", NULL, NULL },
1193 [OPT_FIRST_COL] = { "first-column", "First column", "The column number of the first channel (multi-col. mode)", NULL, NULL },
1194 [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 },
1195 [OPT_START] = { "startline", "Start line", "The line number at which to start processing samples (>= 1)", NULL, NULL },
1196 [OPT_MAX] = ALL_ZERO,
1199 static const struct sr_option *get_options(void)
1203 if (!options[0].def) {
1204 options[OPT_COL_FMTS].def = g_variant_ref_sink(g_variant_new_string(""));
1205 options[OPT_SINGLE_COL].def = g_variant_ref_sink(g_variant_new_uint32(0));
1206 options[OPT_NUM_LOGIC].def = g_variant_ref_sink(g_variant_new_uint32(0));
1207 options[OPT_DELIM].def = g_variant_ref_sink(g_variant_new_string(","));
1208 options[OPT_FORMAT].def = g_variant_ref_sink(g_variant_new_string("bin"));
1210 l = g_slist_append(l, g_variant_ref_sink(g_variant_new_string("bin")));
1211 l = g_slist_append(l, g_variant_ref_sink(g_variant_new_string("hex")));
1212 l = g_slist_append(l, g_variant_ref_sink(g_variant_new_string("oct")));
1213 options[OPT_FORMAT].values = l;
1214 options[OPT_COMMENT].def = g_variant_ref_sink(g_variant_new_string(";"));
1215 options[OPT_RATE].def = g_variant_ref_sink(g_variant_new_uint64(0));
1216 options[OPT_FIRST_COL].def = g_variant_ref_sink(g_variant_new_uint32(1));
1217 options[OPT_HEADER].def = g_variant_ref_sink(g_variant_new_boolean(FALSE));
1218 options[OPT_START].def = g_variant_ref_sink(g_variant_new_uint32(1));
1224 SR_PRIV struct sr_input_module input_csv = {
1227 .desc = "Comma-separated values",
1228 .exts = (const char*[]){"csv", NULL},
1229 .options = get_options,