X-Git-Url: https://sigrok.org/gitweb/?a=blobdiff_plain;f=src%2Finput%2Fcsv.c;h=e3e609917ab5c99cb1e9a77aa5bc648771fa2f1c;hb=1a920e33fefffe474a8949fccb9f604dfef8401b;hp=b8574d3e0542efee8233d4cd9571869075da13f6;hpb=c6aa9870b4e44c94621ea124898b9b860acf86ce;p=libsigrok.git

diff --git a/src/input/csv.c b/src/input/csv.c
index b8574d3e..e3e60991 100644
--- a/src/input/csv.c
+++ b/src/input/csv.c
@@ -2,6 +2,7 @@
  * This file is part of the libsigrok project.
  *
  * Copyright (C) 2013 Marc Schink <sigrok-dev@marcschink.de>
+ * Copyright (C) 2019 Gerhard Sittig <gerhard.sittig@gmx.net>
  *
  * This program is free software: you can redistribute it and/or modify
  * it under the terms of the GNU General Public License as published by
@@ -25,6 +26,7 @@
 
 #include <libsigrok/libsigrok.h>
 #include "libsigrok-internal.h"
+#include "scpi.h"	/* String un-quote for channel name from header line. */
 
 #define LOG_PREFIX "input/csv"
 
@@ -106,9 +108,31 @@
 
 /* Single column formats. */
 enum single_col_format {
-	FORMAT_BIN,
-	FORMAT_HEX,
-	FORMAT_OCT,
+	FORMAT_NONE,	/* Ignore this column. */
+	FORMAT_BIN,	/* Bin digits for a set of bits (or just one bit). */
+	FORMAT_HEX,	/* Hex digits for a set of bits. */
+	FORMAT_OCT,	/* Oct digits for a set of bits. */
+};
+
+static const char *col_format_text[] = {
+	[FORMAT_NONE] = "unknown",
+	[FORMAT_BIN] = "binary",
+	[FORMAT_HEX] = "hexadecimal",
+	[FORMAT_OCT] = "octal",
+};
+
+static const char col_format_char[] = {
+	[FORMAT_NONE] = '?',
+	[FORMAT_BIN] = 'b',
+	[FORMAT_HEX] = 'x',
+	[FORMAT_OCT] = 'o',
+};
+
+struct column_details {
+	size_t col_nr;
+	enum single_col_format text_format;
+	size_t channel_offset;
+	size_t channel_count;
 };
 
 struct context {
@@ -116,37 +140,21 @@ struct context {
 
 	/* Current selected samplerate. */
 	uint64_t samplerate;
+	gboolean samplerate_sent;
 
-	/* Number of channels. */
-	size_t num_channels;
+	/* Number of logic channels. */
+	size_t logic_channels;
 
-	/* Column delimiter character(s). */
+	/* Column delimiter (actually separator), comment leader, EOL sequence. */
 	GString *delimiter;
-
-	/* Comment prefix character(s). */
 	GString *comment;
-
-	/* Termination character(s) used in current stream. */
 	char *termination;
 
-	/* Determines if sample data is stored in multiple columns. */
-	gboolean multi_column_mode;
-
-	/* Column number of the sample data in single column mode. */
-	size_t single_column;
-
-	/*
-	 * Number of the first column to parse. Equivalent to the number of the
-	 * first channel in multi column mode and the single column number in
-	 * single column mode.
-	 */
-	size_t first_column;
-
-	/*
-	 * Column number of the first channel in multi column mode and position of
-	 * the bit for the first channel in single column mode.
-	 */
-	size_t first_channel;
+	/* Format specs for input columns, and processing state. */
+	size_t column_seen_count;
+	const char *column_formats;
+	size_t column_want_count;
+	struct column_details *column_details;
 
 	/* Line number to start processing. */
 	size_t start_line;
@@ -155,10 +163,8 @@ struct context {
 	 * Determines if the first line should be treated as header and used for
 	 * channel names in multi column mode.
 	 */
-	gboolean header;
-
-	/* Format sample data is stored in single column mode. */
-	enum single_col_format format;
+	gboolean use_header;
+	gboolean header_seen;
 
 	size_t sample_unit_size;	/**!< Byte count for a single sample. */
 	uint8_t *sample_buffer;		/**!< Buffer for a single sample. */
@@ -174,133 +180,280 @@ struct context {
 	GSList *prev_sr_channels;
 };
 
-static void strip_comment(char *buf, const GString *prefix)
+/*
+ * Primitive operations to handle sample sets:
+ * - Keep a buffer for datafeed submission, capable of holding many
+ *   samples (reduces call overhead, improves throughput).
+ * - Have a "current sample set" pointer reference one position in that
+ *   large samples buffer.
+ * - Clear the current sample set before text line inspection, then set
+ *   the bits which are found active in the current line of text input.
+ *   Phrase the API such that call sites can be kept simple. Advance to
+ *   the next sample set between lines, flush the larger buffer as needed
+ *   (when it is full, or upon EOF).
+ */
+
+static void clear_logic_samples(struct context *inc)
 {
-	char *ptr;
+	inc->sample_buffer = &inc->datafeed_buffer[inc->datafeed_buf_fill];
+	memset(inc->sample_buffer, 0, inc->sample_unit_size);
+}
 
-	if (!prefix->len)
+static void set_logic_level(struct context *inc, size_t ch_idx, int on)
+{
+	size_t byte_idx, bit_idx;
+	uint8_t bit_mask;
+
+	if (ch_idx >= inc->logic_channels)
+		return;
+	if (!on)
 		return;
 
-	if ((ptr = strstr(buf, prefix->str)))
-		*ptr = '\0';
+	byte_idx = ch_idx / 8;
+	bit_idx = ch_idx % 8;
+	bit_mask = 1 << bit_idx;
+	inc->sample_buffer[byte_idx] |= bit_mask;
 }
 
-static int parse_binstr(const char *str, struct context *inc)
+static int flush_logic_samples(const struct sr_input *in)
 {
-	gsize i, j, length;
+	struct context *inc;
+	struct sr_datafeed_packet packet;
+	struct sr_datafeed_meta meta;
+	struct sr_config *src;
+	uint64_t samplerate;
+	struct sr_datafeed_logic logic;
+	int rc;
 
-	length = strlen(str);
+	inc = in->priv;
+	if (!inc->datafeed_buf_fill)
+		return SR_OK;
 
-	if (!length) {
-		sr_err("Column %zu in line %zu is empty.", inc->single_column,
-			inc->line_number);
-		return SR_ERR;
+	if (inc->samplerate && !inc->samplerate_sent) {
+		packet.type = SR_DF_META;
+		packet.payload = &meta;
+		samplerate = inc->samplerate;
+		src = sr_config_new(SR_CONF_SAMPLERATE, g_variant_new_uint64(samplerate));
+		meta.config = g_slist_append(NULL, src);
+		sr_session_send(in->sdi, &packet);
+		g_slist_free(meta.config);
+		sr_config_free(src);
+		inc->samplerate_sent = TRUE;
 	}
 
-	/* Clear buffer in order to set bits only. */
-	memset(inc->sample_buffer, 0, inc->sample_unit_size);
-
-	i = inc->first_channel;
+	memset(&packet, 0, sizeof(packet));
+	memset(&logic, 0, sizeof(logic));
+	packet.type = SR_DF_LOGIC;
+	packet.payload = &logic;
+	logic.unitsize = inc->sample_unit_size;
+	logic.length = inc->datafeed_buf_fill;
+	logic.data = inc->datafeed_buffer;
 
-	for (j = 0; i < length && j < inc->num_channels; i++, j++) {
-		if (str[length - i - 1] == '1') {
-			inc->sample_buffer[j / 8] |= (1 << (j % 8));
-		} else if (str[length - i - 1] != '0') {
-			sr_err("Invalid value '%s' in column %zu in line %zu.",
-				str, inc->single_column, inc->line_number);
-			return SR_ERR;
-		}
-	}
+	rc = sr_session_send(in->sdi, &packet);
+	if (rc != SR_OK)
+		return rc;
 
+	inc->datafeed_buf_fill = 0;
 	return SR_OK;
 }
 
-static int parse_hexstr(const char *str, struct context *inc)
+static int queue_logic_samples(const struct sr_input *in)
 {
-	gsize i, j, k, length;
-	uint8_t value;
-	char c;
+	struct context *inc;
+	int rc;
 
-	length = strlen(str);
+	inc = in->priv;
+	if (!inc->logic_channels)
+		return SR_OK;
 
-	if (!length) {
-		sr_err("Column %zu in line %zu is empty.", inc->single_column,
-			inc->line_number);
-		return SR_ERR;
+	inc->datafeed_buf_fill += inc->sample_unit_size;
+	if (inc->datafeed_buf_fill == inc->datafeed_buf_size) {
+		rc = flush_logic_samples(in);
+		if (rc != SR_OK)
+			return rc;
 	}
+	return SR_OK;
+}
 
-	/* Clear buffer in order to set bits only. */
-	memset(inc->sample_buffer, 0, inc->sample_unit_size);
+/* Helpers for "column processing". */
 
-	/* Calculate the position of the first hexadecimal digit. */
-	i = inc->first_channel / 4;
+static int split_column_format(const char *spec,
+	size_t *column_count, enum single_col_format *format, size_t *bit_count)
+{
+	size_t count;
+	char *endp, format_char;
+	enum single_col_format format_code;
 
-	for (j = 0; i < length && j < inc->num_channels; i++) {
-		c = str[length - i - 1];
+	if (!spec || !*spec)
+		return SR_ERR_ARG;
 
-		if (!g_ascii_isxdigit(c)) {
-			sr_err("Invalid value '%s' in column %zu in line %zu.",
-				str, inc->single_column, inc->line_number);
-			return SR_ERR;
-		}
+	/* Get the (optional, decimal, default 1) column count. Accept '*'. */
+	endp = NULL;
+	if (*spec == '*') {
+		count = 0;
+		endp = (char *)&spec[1];
+	} else {
+		count = strtoul(spec, &endp, 10);
+	}
+	if (!endp)
+		return SR_ERR_ARG;
+	if (endp == spec)
+		count = 1;
+	if (column_count)
+		*column_count = count;
+	spec = endp;
+
+	/* Get the (mandatory, single letter) type spec (-/xob/l). */
+	format_char = *spec++;
+	switch (format_char) {
+	case '-':	/* Might conflict with number-parsing. */
+	case '/':
+		format_char = '-';
+		format_code = FORMAT_NONE;
+		break;
+	case 'x':
+		format_code = FORMAT_HEX;
+		break;
+	case 'o':
+		format_code = FORMAT_OCT;
+		break;
+	case 'b':
+	case 'l':
+		format_code = FORMAT_BIN;
+		break;
+	default:	/* includes NUL */
+		return SR_ERR_ARG;
+	}
+	if (format)
+		*format = format_code;
 
-		value = g_ascii_xdigit_value(c);
+	/* Get the (optional, decimal, default 1) bit count. */
+	endp = NULL;
+	count = strtoul(spec, &endp, 10);
+	if (!endp)
+		return SR_ERR_ARG;
+	if (endp == spec)
+		count = 1;
+	if (format_char == '-')
+		count = 0;
+	if (format_char == 'l')
+		count = 1;
+	if (bit_count)
+		*bit_count = count;
+	spec = endp;
+
+	/* Input spec must have been exhausted. */
+	if (*spec)
+		return SR_ERR_ARG;
 
-		k = (inc->first_channel + j) % 4;
+	return SR_OK;
+}
 
-		for (; j < inc->num_channels && k < 4; k++) {
-			if (value & (1 << k))
-				inc->sample_buffer[j / 8] |= (1 << (j % 8));
+static int make_column_details_from_format(struct context *inc,
+	const char *column_format)
+{
+	char **formats, *format;
+	size_t format_count, column_count, bit_count;
+	size_t auto_column_count;
+	size_t format_idx, c, b, column_idx, channel_idx;
+	enum single_col_format f;
+	struct column_details *detail;
+	int ret;
 
-			j++;
+	/* Split the input spec, count involved columns and bits. */
+	formats = g_strsplit(column_format, ",", 0);
+	if (!formats) {
+		sr_err("Cannot parse columns format %s (comma split).", column_format);
+		return SR_ERR_ARG;
+	}
+	format_count = g_strv_length(formats);
+	if (!format_count) {
+		sr_err("Cannot parse columns format %s (field count).", column_format);
+		g_strfreev(formats);
+		return SR_ERR_ARG;
+	}
+	column_count = bit_count = 0;
+	auto_column_count = 0;
+	for (format_idx = 0; format_idx < format_count; format_idx++) {
+		format = formats[format_idx];
+		ret = split_column_format(format, &c, &f, &b);
+		sr_dbg("fmt %s -> %zu cols, %s fmt, %zu bits, rc %d", format, c, col_format_text[f], b, ret);
+		if (ret != SR_OK) {
+			sr_err("Cannot parse columns format %s (field split, %s).", column_format, format);
+			g_strfreev(formats);
+			return SR_ERR_ARG;
+		}
+		if (f && !c) {
+			/* User requested "auto-count", must be last format. */
+			if (formats[format_idx + 1]) {
+				sr_err("Auto column count must be last format field.");
+				g_strfreev(formats);
+				return SR_ERR_ARG;
+			}
+			auto_column_count = inc->column_seen_count - column_count;
+			c = auto_column_count;
+		}
+		column_count += c;
+		bit_count += c * b;
+	}
+	sr_dbg("Column format %s -> %zu columns, %zu logic channels.",
+		column_format, column_count, bit_count);
+
+	/* Allocate and fill in "column processing" details. */
+	inc->column_want_count = column_count;
+	inc->column_details = g_malloc0_n(column_count, sizeof(inc->column_details[0]));
+	column_idx = channel_idx = 0;
+	for (format_idx = 0; format_idx < format_count; format_idx++) {
+		format = formats[format_idx];
+		(void)split_column_format(format, &c, &f, &b);
+		if (f && !c)
+			c = auto_column_count;
+		while (c-- > 0) {
+			detail = &inc->column_details[column_idx++];
+			detail->col_nr = column_idx;
+			detail->text_format = f;
+			if (detail->text_format) {
+				detail->channel_offset = channel_idx;
+				detail->channel_count = b;
+				channel_idx += b;
+			}
+			sr_dbg("detail -> col %zu, fmt %s, ch off/cnt %zu/%zu",
+				detail->col_nr, col_format_text[detail->text_format],
+				detail->channel_offset, detail->channel_count);
 		}
 	}
+	inc->logic_channels = channel_idx;
+	g_strfreev(formats);
 
 	return SR_OK;
 }
 
-static int parse_octstr(const char *str, struct context *inc)
+static const struct column_details *lookup_column_details(struct context *inc, size_t nr)
 {
-	gsize i, j, k, length;
-	uint8_t value;
-	char c;
-
-	length = strlen(str);
-
-	if (!length) {
-		sr_err("Column %zu in line %zu is empty.", inc->single_column,
-			inc->line_number);
-		return SR_ERR;
-	}
-
-	/* Clear buffer in order to set bits only. */
-	memset(inc->sample_buffer, 0, inc->sample_unit_size);
-
-	/* Calculate the position of the first octal digit. */
-	i = inc->first_channel / 3;
-
-	for (j = 0; i < length && j < inc->num_channels; i++) {
-		c = str[length - i - 1];
-
-		if (c < '0' || c > '7') {
-			sr_err("Invalid value '%s' in column %zu in line %zu.",
-				str, inc->single_column, inc->line_number);
-			return SR_ERR;
-		}
+	if (!inc || !inc->column_details)
+		return NULL;
+	if (!nr || nr > inc->column_want_count)
+		return NULL;
+	return &inc->column_details[nr - 1];
+}
 
-		value = g_ascii_xdigit_value(c);
+/*
+ * Primitive operations for text input: Strip comments off text lines.
+ * Split text lines into columns. Process input text for individual
+ * columns.
+ */
 
-		k = (inc->first_channel + j) % 3;
+static void strip_comment(char *buf, const GString *prefix)
+{
+	char *ptr;
 
-		for (; j < inc->num_channels && k < 3; k++) {
-			if (value & (1 << k))
-				inc->sample_buffer[j / 8] |= (1 << (j % 8));
+	if (!prefix->len)
+		return;
 
-			j++;
-		}
+	if ((ptr = strstr(buf, prefix->str))) {
+		*ptr = '\0';
+		g_strstrip(buf);
 	}
-
-	return SR_OK;
 }
 
 /**
@@ -308,177 +461,180 @@ static int parse_octstr(const char *str, struct context *inc)
  *
  * @param[in] buf	The input text line to split.
  * @param[in] inc	The input module's context.
- * @param[in] max_cols	The maximum column count, negative to get all of them.
  *
  * @returns An array of strings, representing the columns' text.
+ *
+ * This routine splits a text line on previously determined separators.
  */
-static char **parse_line(char *buf, struct context *inc, ssize_t max_cols)
+static char **split_line(char *buf, struct context *inc)
 {
-	const char *str, *remainder;
-	GSList *list, *l;
-	char **columns;
-	char *column;
-	gsize n, k;
-
-	n = 0;
-	k = 0;
-	list = NULL;
-
-	remainder = buf;
-	str = strstr(remainder, inc->delimiter->str);
-
-	while (str && max_cols) {
-		if (n >= inc->first_column) {
-			column = g_strndup(remainder, str - remainder);
-			list = g_slist_prepend(list, g_strstrip(column));
-
-			max_cols--;
-			k++;
-		}
-
-		remainder = str + inc->delimiter->len;
-		str = strstr(remainder, inc->delimiter->str);
-		n++;
-	}
-
-	if (buf[0] && max_cols && n >= inc->first_column) {
-		column = g_strdup(remainder);
-		list = g_slist_prepend(list, g_strstrip(column));
-		k++;
-	}
-
-	if (!(columns = g_try_new(char *, k + 1)))
-		return NULL;
-
-	columns[k--] = NULL;
-
-	for (l = list; l; l = l->next)
-		columns[k--] = l->data;
-
-	g_slist_free(list);
-
-	return columns;
+	return g_strsplit(buf, inc->delimiter->str, 0);
 }
 
-static int parse_multi_columns(char **columns, struct context *inc)
+/**
+ * @brief Parse a multi-bit field into several logic channels.
+ *
+ * @param[in] column	The input text, a run of bin/hex/oct digits.
+ * @param[in] inc	The input module's context.
+ * @param[in] details	The column processing details.
+ *
+ * @retval SR_OK	Success.
+ * @retval SR_ERR	Invalid input data (empty, or format error).
+ *
+ * This routine modifies the logic levels in the current sample set,
+ * based on the text input and a user provided format spec.
+ */
+static int parse_logic(const char *column, struct context *inc,
+	const struct column_details *details)
 {
-	gsize i;
-	char *column;
+	size_t length, ch_rem, ch_idx, ch_inc;
+	const char *rdptr;
+	char c;
+	gboolean valid;
+	const char *type_text;
+	uint8_t bits;
 
-	/* Clear buffer in order to set bits only. */
-	memset(inc->sample_buffer, 0, inc->sample_unit_size);
+	/*
+	 * Prepare to read the digits from the text end towards the start.
+	 * A digit corresponds to a variable number of channels (depending
+	 * on the value's radix). Prepare the mapping of text digits to
+	 * (a number of) logic channels.
+	 */
+	length = strlen(column);
+	if (!length) {
+		sr_err("Column %zu in line %zu is empty.", details->col_nr,
+			inc->line_number);
+		return SR_ERR;
+	}
+	rdptr = &column[length];
+	ch_idx = details->channel_offset;
+	ch_rem = details->channel_count;
 
-	for (i = 0; i < inc->num_channels; i++) {
-		column = columns[i];
-		if (column[0] == '1') {
-			inc->sample_buffer[i / 8] |= (1 << (i % 8));
-		} else if (!strlen(column)) {
-			sr_err("Column %zu in line %zu is empty.",
-				inc->first_channel + i, inc->line_number);
+	/*
+	 * Get another digit and derive up to four logic channels' state from
+	 * it. Make sure to not process more bits than the column has channels
+	 * associated with it.
+	 */
+	while (rdptr > column && ch_rem) {
+		/* Check for valid digits according to the input radix. */
+		c = *(--rdptr);
+		switch (details->text_format) {
+		case FORMAT_BIN:
+			valid = g_ascii_isxdigit(c) && c < '2';
+			ch_inc = 1;
+			break;
+		case FORMAT_OCT:
+			valid = g_ascii_isxdigit(c) && c < '8';
+			ch_inc = 3;
+			break;
+		case FORMAT_HEX:
+			valid = g_ascii_isxdigit(c);
+			ch_inc = 4;
+			break;
+		default:
+			valid = FALSE;
+			break;
+		}
+		if (!valid) {
+			type_text = col_format_text[details->text_format];
+			sr_err("Invalid text '%s' in %s type column %zu in line %zu.",
+				column, type_text, details->col_nr, inc->line_number);
 			return SR_ERR;
-		} else if (column[0] != '0') {
-			sr_err("Invalid value '%s' in column %zu in line %zu.",
-				column, inc->first_channel + i,
-				inc->line_number);
+		}
+		/* Use the digit's bits for logic channels' data. */
+		bits = g_ascii_xdigit_value(c);
+		switch (details->text_format) {
+		case FORMAT_HEX:
+			if (ch_rem >= 4) {
+				ch_rem--;
+				set_logic_level(inc, ch_idx + 3, bits & (1 << 3));
+			}
+			/* FALLTHROUGH */
+		case FORMAT_OCT:
+			if (ch_rem >= 3) {
+				ch_rem--;
+				set_logic_level(inc, ch_idx + 2, bits & (1 << 2));
+			}
+			if (ch_rem >= 2) {
+				ch_rem--;
+				set_logic_level(inc, ch_idx + 1, bits & (1 << 1));
+			}
+			/* FALLTHROUGH */
+		case FORMAT_BIN:
+			ch_rem--;
+			set_logic_level(inc, ch_idx + 0, bits & (1 << 0));
+			break;
+		case FORMAT_NONE:
+			/* ShouldNotHappen(TM), but silences compiler warning. */
 			return SR_ERR;
 		}
+		ch_idx += ch_inc;
 	}
+	/*
+	 * TODO Determine whether the availability of extra input data
+	 * for unhandled logic channels is worth warning here. In this
+	 * implementation users are in control, and can have the more
+	 * significant bits ignored (which can be considered a feature
+	 * and not really a limitation).
+	 */
 
 	return SR_OK;
 }
 
-static int parse_single_column(const char *column, struct context *inc)
-{
-	int res;
-
-	res = SR_ERR;
-
-	switch (inc->format) {
-	case FORMAT_BIN:
-		res = parse_binstr(column, inc);
-		break;
-	case FORMAT_HEX:
-		res = parse_hexstr(column, inc);
-		break;
-	case FORMAT_OCT:
-		res = parse_octstr(column, inc);
-		break;
-	}
-
-	return res;
-}
-
-static int flush_samples(const struct sr_input *in)
+/**
+ * @brief Parse routine which ignores the input text.
+ *
+ * This routine exists to unify dispatch code paths, mapping input file
+ * columns' data types to their respective parse routines.
+ */
+static int parse_ignore(const char *column, struct context *inc,
+	const struct column_details *details)
 {
-	struct context *inc;
-	struct sr_datafeed_packet packet;
-	struct sr_datafeed_logic logic;
-	int rc;
-
-	inc = in->priv;
-	if (!inc->datafeed_buf_fill)
-		return SR_OK;
-
-	memset(&packet, 0, sizeof(packet));
-	memset(&logic, 0, sizeof(logic));
-	packet.type = SR_DF_LOGIC;
-	packet.payload = &logic;
-	logic.unitsize = inc->sample_unit_size;
-	logic.length = inc->datafeed_buf_fill;
-	logic.data = inc->datafeed_buffer;
-
-	rc = sr_session_send(in->sdi, &packet);
-	if (rc != SR_OK)
-		return rc;
-
-	inc->datafeed_buf_fill = 0;
+	(void)column;
+	(void)inc;
+	(void)details;
 	return SR_OK;
 }
 
-static int queue_samples(const struct sr_input *in)
-{
-	struct context *inc;
-	int rc;
-
-	inc = in->priv;
+typedef int (*col_parse_cb)(const char *column, struct context *inc,
+	const struct column_details *details);
 
-	inc->datafeed_buf_fill += inc->sample_unit_size;
-	if (inc->datafeed_buf_fill == inc->datafeed_buf_size) {
-		rc = flush_samples(in);
-		if (rc != SR_OK)
-			return rc;
-	}
-	inc->sample_buffer = &inc->datafeed_buffer[inc->datafeed_buf_fill];
-	return SR_OK;
-}
+static const col_parse_cb col_parse_funcs[] = {
+	[FORMAT_NONE] = parse_ignore,
+	[FORMAT_BIN] = parse_logic,
+	[FORMAT_OCT] = parse_logic,
+	[FORMAT_HEX] = parse_logic,
+};
 
 static int init(struct sr_input *in, GHashTable *options)
 {
 	struct context *inc;
+	size_t single_column, first_column, logic_channels;
 	const char *s;
+	enum single_col_format format;
+	char format_char;
 
-	in->sdi = g_malloc0(sizeof(struct sr_dev_inst));
-	in->priv = inc = g_malloc0(sizeof(struct context));
+	in->sdi = g_malloc0(sizeof(*in->sdi));
+	in->priv = inc = g_malloc0(sizeof(*inc));
 
-	inc->single_column = g_variant_get_uint32(g_hash_table_lookup(options, "single-column"));
-	inc->multi_column_mode = inc->single_column == 0;
+	single_column = g_variant_get_uint32(g_hash_table_lookup(options, "single-column"));
 
-	inc->num_channels = g_variant_get_uint32(g_hash_table_lookup(options, "numchannels"));
+	logic_channels = g_variant_get_uint32(g_hash_table_lookup(options, "numchannels"));
 
 	inc->delimiter = g_string_new(g_variant_get_string(
 			g_hash_table_lookup(options, "delimiter"), NULL));
-	if (inc->delimiter->len == 0) {
-		sr_err("Delimiter must be at least one character.");
+	if (!inc->delimiter->len) {
+		sr_err("Column delimiter cannot be empty.");
 		return SR_ERR_ARG;
 	}
 
 	s = g_variant_get_string(g_hash_table_lookup(options, "format"), NULL);
-	if (!g_ascii_strncasecmp(s, "bin", 3)) {
-		inc->format = FORMAT_BIN;
-	} else if (!g_ascii_strncasecmp(s, "hex", 3)) {
-		inc->format = FORMAT_HEX;
-	} else if (!g_ascii_strncasecmp(s, "oct", 3)) {
-		inc->format = FORMAT_OCT;
+	if (g_ascii_strncasecmp(s, "bin", 3) == 0) {
+		format = FORMAT_BIN;
+	} else if (g_ascii_strncasecmp(s, "hex", 3) == 0) {
+		format = FORMAT_HEX;
+	} else if (g_ascii_strncasecmp(s, "oct", 3) == 0) {
+		format = FORMAT_OCT;
 	} else {
 		sr_err("Invalid format: '%s'", s);
 		return SR_ERR_ARG;
@@ -487,17 +643,21 @@ static int init(struct sr_input *in, GHashTable *options)
 	inc->comment = g_string_new(g_variant_get_string(
 			g_hash_table_lookup(options, "comment"), NULL));
 	if (g_string_equal(inc->comment, inc->delimiter)) {
-		/* That's never going to work. Likely the result of the user
-		 * setting the delimiter to ; -- the default comment. Clearing
-		 * the comment setting will work in that case. */
+		/*
+		 * Using the same sequence as comment leader and column
+		 * delimiter won't work. The user probably specified ';'
+		 * as the column delimiter but did not adjust the comment
+		 * leader. Try DWIM, drop comment strippin support here.
+		 */
+		sr_warn("Comment leader and column delimiter conflict, disabling comment support.");
 		g_string_truncate(inc->comment, 0);
 	}
 
 	inc->samplerate = g_variant_get_uint64(g_hash_table_lookup(options, "samplerate"));
 
-	inc->first_channel = g_variant_get_uint32(g_hash_table_lookup(options, "first-channel"));
+	first_column = g_variant_get_uint32(g_hash_table_lookup(options, "first-column"));
 
-	inc->header = g_variant_get_boolean(g_hash_table_lookup(options, "header"));
+	inc->use_header = g_variant_get_boolean(g_hash_table_lookup(options, "header"));
 
 	inc->start_line = g_variant_get_uint32(g_hash_table_lookup(options, "startline"));
 	if (inc->start_line < 1) {
@@ -505,14 +665,48 @@ static int init(struct sr_input *in, GHashTable *options)
 		return SR_ERR_ARG;
 	}
 
-	if (inc->multi_column_mode)
-		inc->first_column = inc->first_channel;
-	else
-		inc->first_column = inc->single_column;
-
-	if (!inc->multi_column_mode && !inc->num_channels) {
-		sr_err("Number of channels needs to be specified in single column mode.");
-		return SR_ERR_ARG;
+	/*
+	 * Scan flexible, to get prefered format specs which describe
+	 * the input file's data formats. As well as some simple specs
+	 * for backwards compatibility and user convenience.
+	 *
+	 * This logic ends up with a copy of the format string, either
+	 * user provided or internally derived. Actual creation of the
+	 * column processing details gets deferred until the first line
+	 * of input data was seen. To support automatic determination of
+	 * e.g. channel counts from column counts.
+	 */
+	s = g_variant_get_string(g_hash_table_lookup(options, "column-formats"), NULL);
+	if (s && *s) {
+		inc->column_formats = g_strdup(s);
+		sr_dbg("User specified column-formats: %s.", s);
+	} else if (single_column && logic_channels) {
+		format_char = col_format_char[format];
+		if (single_column == 1) {
+			inc->column_formats = g_strdup_printf("%c%zu",
+				format_char, logic_channels);
+		} else {
+			inc->column_formats = g_strdup_printf("%zu-,%c%zu",
+				single_column - 1,
+				format_char, logic_channels);
+		}
+		sr_dbg("Backwards compat single-column, col %zu, fmt %s, bits %zu -> %s.",
+			single_column, col_format_text[format], logic_channels,
+			inc->column_formats);
+	} else if (!single_column) {
+		if (first_column > 1) {
+			inc->column_formats = g_strdup_printf("%zu-,%zul",
+				first_column - 1, logic_channels);
+		} else {
+			inc->column_formats = g_strdup_printf("%zul",
+				logic_channels);
+		}
+		sr_dbg("Backwards compat multi-column, col %zu, chans %zu -> %s.",
+			first_column, logic_channels,
+			inc->column_formats);
+	} else {
+		sr_warn("Unknown or unsupported format spec, assuming trivial multi-column.");
+		inc->column_formats = g_strdup("*l");
 	}
 
 	return SR_OK;
@@ -577,22 +771,27 @@ static int initial_parse(const struct sr_input *in, GString *buf)
 {
 	struct context *inc;
 	GString *channel_name;
-	size_t num_columns, i;
-	size_t line_number, l;
+	size_t num_columns, ch_idx, ch_name_idx, col_idx, col_nr;
+	size_t line_number, line_idx;
 	int ret;
 	char **lines, *line, **columns, *column;
+	const char *col_caption;
+	gboolean got_caption;
+	const struct column_details *detail;
 
 	ret = SR_OK;
 	inc = in->priv;
 	columns = NULL;
 
 	line_number = 0;
-	lines = g_strsplit_set(buf->str, delim_set, 0);
-	for (l = 0; lines[l]; l++) {
+	if (inc->termination)
+		lines = g_strsplit(buf->str, inc->termination, 0);
+	else
+		lines = g_strsplit_set(buf->str, delim_set, 0);
+	for (line_idx = 0; (line = lines[line_idx]); line_idx++) {
 		line_number++;
-		line = lines[l];
 		if (inc->start_line > line_number) {
-			sr_spew("Line %zu skipped.", line_number);
+			sr_spew("Line %zu skipped (before start).", line_number);
 			continue;
 		}
 		if (line[0] == '\0') {
@@ -608,63 +807,86 @@ static int initial_parse(const struct sr_input *in, GString *buf)
 		/* Reached first proper line. */
 		break;
 	}
-	if (!lines[l]) {
+	if (!line) {
 		/* Not enough data for a proper line yet. */
 		ret = SR_ERR_NA;
 		goto out;
 	}
 
-	/*
-	 * In order to determine the number of columns parse the current line
-	 * without limiting the number of columns.
-	 */
-	columns = parse_line(line, inc, -1);
+	/* See how many columns the current line has. */
+	columns = split_line(line, inc);
 	if (!columns) {
 		sr_err("Error while parsing line %zu.", line_number);
 		ret = SR_ERR;
 		goto out;
 	}
 	num_columns = g_strv_length(columns);
-
-	/* Ensure that the first column is not out of bounds. */
 	if (!num_columns) {
-		sr_err("Column %zu in line %zu is out of bounds.",
-			inc->first_column, line_number);
+		sr_err("Error while parsing line %zu.", line_number);
 		ret = SR_ERR;
 		goto out;
 	}
+	sr_dbg("DIAG Got %zu columns in text line: %s.", num_columns, line);
 
-	if (inc->multi_column_mode) {
-		/*
-		 * Detect the number of channels in multi column mode
-		 * automatically if not specified.
-		 */
-		if (!inc->num_channels) {
-			inc->num_channels = num_columns;
-			sr_dbg("Number of auto-detected channels: %zu.",
-				inc->num_channels);
-		}
+	/*
+	 * Track the observed number of columns in the input file. Do
+	 * process the previously gathered columns format spec now that
+	 * automatic channel count can be dealt with.
+	 */
+	inc->column_seen_count = num_columns;
+	ret = make_column_details_from_format(inc, inc->column_formats);
+	if (ret != SR_OK) {
+		sr_err("Cannot parse columns format using line %zu.", line_number);
+		goto out;
+	}
 
-		/*
-		 * Ensure that the number of channels does not exceed the number
-		 * of columns in multi column mode.
-		 */
-		if (num_columns < inc->num_channels) {
-			sr_err("Not enough columns for desired number of channels in line %zu.",
-				line_number);
-			ret = SR_ERR;
-			goto out;
-		}
+	/*
+	 * Assume all lines have equal length (column count). Bail out
+	 * early on suspicious or insufficient input data (check input
+	 * which became available here against previous user specs or
+	 * auto-determined properties, regardless of layout variant).
+	 */
+	if (num_columns < inc->column_want_count) {
+		sr_err("Insufficient input text width for desired data amount, got %zu but want %zu columns.",
+			num_columns, inc->column_want_count);
+		ret = SR_ERR;
+		goto out;
 	}
 
+	/*
+	 * Determine channel names. Optionally use text from a header
+	 * line (when requested by the user, and only works in multi
+	 * column mode). In the absence of header text, or in single
+	 * column mode, channels are assigned rather generic names.
+	 *
+	 * Manipulation of the column's caption is acceptable here, the
+	 * header line will never get processed another time.
+	 */
 	channel_name = g_string_sized_new(64);
-	for (i = 0; i < inc->num_channels; i++) {
-		column = columns[i];
-		if (inc->header && inc->multi_column_mode && column[0] != '\0')
-			g_string_assign(channel_name, column);
-		else
-			g_string_printf(channel_name, "%zu", i);
-		sr_channel_new(in->sdi, i, SR_CHANNEL_LOGIC, TRUE, channel_name->str);
+	for (col_idx = 0; col_idx < inc->column_want_count; col_idx++) {
+
+		col_nr = col_idx + 1;
+		detail = lookup_column_details(inc, col_nr);
+		if (detail->text_format == FORMAT_NONE)
+			continue;
+		column = columns[col_idx];
+		col_caption = sr_scpi_unquote_string(column);
+		got_caption = inc->use_header && *col_caption;
+		sr_dbg("DIAG col %zu, ch count %zu, text %s.",
+			col_nr, detail->channel_count, col_caption);
+		for (ch_idx = 0; ch_idx < detail->channel_count; ch_idx++) {
+			ch_name_idx = detail->channel_offset + ch_idx;
+			if (got_caption && detail->channel_count == 1)
+				g_string_assign(channel_name, col_caption);
+			else if (got_caption)
+				g_string_printf(channel_name, "%s[%zu]",
+					col_caption, ch_idx);
+			else
+				g_string_printf(channel_name, "%zu", ch_name_idx);
+			sr_dbg("DIAG ch idx %zu, name %s.", ch_name_idx, channel_name->str);
+			sr_channel_new(in->sdi, ch_name_idx, SR_CHANNEL_LOGIC, TRUE,
+				channel_name->str);
+		}
 	}
 	g_string_free(channel_name, TRUE);
 	if (!check_header_in_reread(in)) {
@@ -676,15 +898,14 @@ static int initial_parse(const struct sr_input *in, GString *buf)
 	 * Calculate the minimum buffer size to store the set of samples
 	 * of all channels (unit size). Determine a larger buffer size
 	 * for datafeed submission that is a multiple of the unit size.
-	 * Allocate the larger buffer, and have the "sample buffer" point
-	 * to a location within that large buffer.
+	 * Allocate the larger buffer, the "sample buffer" will point
+	 * to a location within that large buffer later.
 	 */
-	inc->sample_unit_size = (inc->num_channels + 7) / 8;
+	inc->sample_unit_size = (inc->logic_channels + 7) / 8;
 	inc->datafeed_buf_size = CHUNK_SIZE;
 	inc->datafeed_buf_size *= inc->sample_unit_size;
 	inc->datafeed_buffer = g_malloc(inc->datafeed_buf_size);
 	inc->datafeed_buf_fill = 0;
-	inc->sample_buffer = &inc->datafeed_buffer[inc->datafeed_buf_fill];
 
 out:
 	if (columns)
@@ -754,40 +975,20 @@ static int initial_receive(const struct sr_input *in)
 
 static int process_buffer(struct sr_input *in, gboolean is_eof)
 {
-	struct sr_datafeed_packet packet;
-	struct sr_datafeed_meta meta;
-	struct sr_config *src;
 	struct context *inc;
 	gsize num_columns;
-	uint64_t samplerate;
-	size_t max_columns, l;
+	size_t line_idx, col_idx, col_nr;
+	const struct column_details *details;
+	col_parse_cb parse_func;
 	int ret;
-	char *p, **lines, *line, **columns;
+	char *p, **lines, *line, **columns, *column;
 
 	inc = in->priv;
 	if (!inc->started) {
 		std_session_send_df_header(in->sdi);
-
-		if (inc->samplerate) {
-			packet.type = SR_DF_META;
-			packet.payload = &meta;
-			samplerate = inc->samplerate;
-			src = sr_config_new(SR_CONF_SAMPLERATE, g_variant_new_uint64(samplerate));
-			meta.config = g_slist_append(NULL, src);
-			sr_session_send(in->sdi, &packet);
-			g_slist_free(meta.config);
-			sr_config_free(src);
-		}
-
 		inc->started = TRUE;
 	}
 
-	/* Limit the number of columns to parse. */
-	if (inc->multi_column_mode)
-		max_columns = inc->num_channels;
-	else
-		max_columns = 1;
-
 	/*
 	 * Consider empty input non-fatal. Keep accumulating input until
 	 * at least one full text line has become available. Grab the
@@ -815,10 +1016,13 @@ static int process_buffer(struct sr_input *in, gboolean is_eof)
 	g_strstrip(in->buf->str);
 
 	ret = SR_OK;
-	lines = g_strsplit_set(in->buf->str, delim_set, 0);
-	for (l = 0; lines[l]; l++) {
+	lines = g_strsplit(in->buf->str, inc->termination, 0);
+	for (line_idx = 0; (line = lines[line_idx]); line_idx++) {
 		inc->line_number++;
-		line = lines[l];
+		if (inc->line_number < inc->start_line) {
+			sr_spew("Line %zu skipped (before start).", inc->line_number);
+			continue;
+		}
 		if (line[0] == '\0') {
 			sr_spew("Blank line %zu skipped.", inc->line_number);
 			continue;
@@ -832,50 +1036,49 @@ static int process_buffer(struct sr_input *in, gboolean is_eof)
 		}
 
 		/* Skip the header line, its content was used as the channel names. */
-		if (inc->header) {
+		if (inc->use_header && !inc->header_seen) {
 			sr_spew("Header line %zu skipped.", inc->line_number);
-			inc->header = FALSE;
+			inc->header_seen = TRUE;
 			continue;
 		}
 
-		columns = parse_line(line, inc, max_columns);
+		/* Split the line into columns, check for minimum length. */
+		columns = split_line(line, inc);
 		if (!columns) {
 			sr_err("Error while parsing line %zu.", inc->line_number);
 			g_strfreev(lines);
 			return SR_ERR;
 		}
 		num_columns = g_strv_length(columns);
-		if (!num_columns) {
-			sr_err("Column %zu in line %zu is out of bounds.",
-				inc->first_column, inc->line_number);
-			g_strfreev(columns);
-			g_strfreev(lines);
-			return SR_ERR;
-		}
-		/*
-		 * Ensure that the number of channels does not exceed the number
-		 * of columns in multi column mode.
-		 */
-		if (inc->multi_column_mode && num_columns < inc->num_channels) {
-			sr_err("Not enough columns for desired number of channels in line %zu.",
-				inc->line_number);
+		if (num_columns < inc->column_want_count) {
+			sr_err("Insufficient column count %zu in line %zu.",
+				num_columns, inc->line_number);
 			g_strfreev(columns);
 			g_strfreev(lines);
 			return SR_ERR;
 		}
 
-		if (inc->multi_column_mode)
-			ret = parse_multi_columns(columns, inc);
-		else
-			ret = parse_single_column(columns[0], inc);
-		if (ret != SR_OK) {
-			g_strfreev(columns);
-			g_strfreev(lines);
-			return SR_ERR;
+		/* Have the columns of the current text line processed. */
+		clear_logic_samples(inc);
+		for (col_idx = 0; col_idx < inc->column_want_count; col_idx++) {
+			column = columns[col_idx];
+			col_nr = col_idx + 1;
+			details = lookup_column_details(inc, col_nr);
+			if (!details || !details->text_format)
+				continue;
+			parse_func = col_parse_funcs[details->text_format];
+			if (!parse_func)
+				continue;
+			ret = parse_func(column, inc, details);
+			if (ret != SR_OK) {
+				g_strfreev(columns);
+				g_strfreev(lines);
+				return SR_ERR;
+			}
 		}
 
-		/* Send sample data to the session bus. */
-		ret = queue_samples(in);
+		/* Send sample data to the session bus (buffered). */
+		ret = queue_logic_samples(in);
 		if (ret != SR_OK) {
 			sr_err("Sending samples failed.");
 			g_strfreev(columns);
@@ -899,7 +1102,7 @@ static int receive(struct sr_input *in, GString *buf)
 	g_string_append_len(in->buf, buf->str, buf->len);
 
 	inc = in->priv;
-	if (!inc->termination) {
+	if (!inc->column_seen_count) {
 		ret = initial_receive(in);
 		if (ret == SR_ERR_NA)
 			/* Not enough data yet. */
@@ -929,7 +1132,7 @@ static int end(struct sr_input *in)
 	if (ret != SR_OK)
 		return ret;
 
-	ret = flush_samples(in);
+	ret = flush_logic_samples(in);
 	if (ret != SR_OK)
 		return ret;
 
@@ -966,26 +1169,28 @@ static int reset(struct sr_input *in)
 }
 
 enum option_index {
+	OPT_COL_FMTS,
 	OPT_SINGLE_COL,
 	OPT_NUM_LOGIC,
 	OPT_DELIM,
 	OPT_FORMAT,
 	OPT_COMMENT,
 	OPT_RATE,
-	OPT_FIRST_LOGIC,
+	OPT_FIRST_COL,
 	OPT_HEADER,
 	OPT_START,
 	OPT_MAX,
 };
 
 static struct sr_option options[] = {
+	[OPT_COL_FMTS] = { "column-formats", "Column format specs", "Specifies text columns data types: comma separated list of [<cols>]<fmt>[<bits>]", NULL, NULL },
 	[OPT_SINGLE_COL] = { "single-column", "Single column", "Enable single-column mode, using the specified column (>= 1); 0: multi-col. mode", NULL, NULL },
 	[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 },
 	[OPT_DELIM] = { "delimiter", "Column delimiter", "The column delimiter (>= 1 characters)", NULL, NULL },
 	[OPT_FORMAT] = { "format", "Data format (single-col. mode)", "The numeric format of the data (single-col. mode): bin, hex, oct", NULL, NULL },
 	[OPT_COMMENT] = { "comment", "Comment character(s)", "The comment prefix character(s)", NULL, NULL },
 	[OPT_RATE] = { "samplerate", "Samplerate (Hz)", "The sample rate (used during capture) in Hz", NULL, NULL },
-	[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 },
+	[OPT_FIRST_COL] = { "first-column", "First column", "The column number of the first channel (multi-col. mode)", NULL, NULL },
 	[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 },
 	[OPT_START] = { "startline", "Start line", "The line number at which to start processing samples (>= 1)", NULL, NULL },
 	[OPT_MAX] = ALL_ZERO,
@@ -996,8 +1201,9 @@ static const struct sr_option *get_options(void)
 	GSList *l;
 
 	if (!options[0].def) {
-		options[OPT_SINGLE_COL].def = g_variant_ref_sink(g_variant_new_int32(0));
-		options[OPT_NUM_LOGIC].def = g_variant_ref_sink(g_variant_new_int32(0));
+		options[OPT_COL_FMTS].def = g_variant_ref_sink(g_variant_new_string(""));
+		options[OPT_SINGLE_COL].def = g_variant_ref_sink(g_variant_new_uint32(0));
+		options[OPT_NUM_LOGIC].def = g_variant_ref_sink(g_variant_new_uint32(0));
 		options[OPT_DELIM].def = g_variant_ref_sink(g_variant_new_string(","));
 		options[OPT_FORMAT].def = g_variant_ref_sink(g_variant_new_string("bin"));
 		l = NULL;
@@ -1007,9 +1213,9 @@ static const struct sr_option *get_options(void)
 		options[OPT_FORMAT].values = l;
 		options[OPT_COMMENT].def = g_variant_ref_sink(g_variant_new_string(";"));
 		options[OPT_RATE].def = g_variant_ref_sink(g_variant_new_uint64(0));
-		options[OPT_FIRST_LOGIC].def = g_variant_ref_sink(g_variant_new_int32(0));
+		options[OPT_FIRST_COL].def = g_variant_ref_sink(g_variant_new_uint32(1));
 		options[OPT_HEADER].def = g_variant_ref_sink(g_variant_new_boolean(FALSE));
-		options[OPT_START].def = g_variant_ref_sink(g_variant_new_int32(1));
+		options[OPT_START].def = g_variant_ref_sink(g_variant_new_uint32(1));
 	}
 
 	return options;