X-Git-Url: https://sigrok.org/gitweb/?a=blobdiff_plain;f=src%2Finput%2Fcsv.c;h=17c253a29d2f29aff5c3cabc6397da1d1fbfecc0;hb=f6dcb3200d12089690976aa6fb25f7071eb4220a;hp=984a58c7d9753ae1d11869d829ad2e5507e9d686;hpb=3be42bc22f8b36599a448273c12a76d3e0f7a940;p=libsigrok.git diff --git a/src/input/csv.c b/src/input/csv.c index 984a58c7..17c253a2 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 + * Copyright (C) 2019 Gerhard Sittig * * 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 @@ -17,15 +18,20 @@ * along with this program. If not, see . */ -#include +#include "config.h" + +#include #include #include -#include + #include #include "libsigrok-internal.h" +#include "scpi.h" /* String un-quote for channel name from header line. */ #define LOG_PREFIX "input/csv" +#define CHUNK_SIZE (4 * 1024 * 1024) + /* * The CSV input module has the following options: * @@ -67,11 +73,59 @@ * than 0. The default line number to start processing is 1. */ +/* + * TODO + * + * - Determine how the text line handling can get improved, regarding + * all of robustness and flexibility and correctness. + * - The current implementation splits on "any run of CR and LF". Which + * translates to: Line numbers are wrong in the presence of empty + * lines in the input stream. See below for an (expensive) fix. + * - Dropping support for CR style end-of-line markers could improve + * the situation a lot. Code could search for and split on LF, and + * trim optional trailing CR. This would result in proper support + * for CRLF (Windows) as well as LF (Unix), and allow for correct + * line number counts. + * - When support for CR-only line termination cannot get dropped, + * then the current implementation is inappropriate. Currently the + * input stream is scanned for the first occurance of either of the + * supported termination styles (which is good). For the remaining + * session a consistent encoding of the text lines is assumed (which + * is acceptable). + * - When line numbers need to be correct and reliable, _and_ the full + * set of previously supported line termination sequences are required, + * and potentially more are to get added for improved compatibility + * with more platforms or generators, then the current approach of + * splitting on runs of termination characters needs to get replaced, + * by the more expensive approach to scan for and count the initially + * determined termination sequence. + * + * - Add support for analog input data? (optional) + * - Needs a syntax first for user specs which channels (columns) are + * logic and which are analog. May need heuristics(?) to guess from + * input data in the absence of user provided specs. + */ + /* Single column formats. */ -enum { - FORMAT_BIN, - FORMAT_HEX, - FORMAT_OCT +enum single_col_format { + 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", +}; + +struct column_details { + size_t col_nr; + enum single_col_format text_format; + size_t channel_offset; + size_t channel_count; }; struct context { @@ -79,9 +133,10 @@ struct context { /* Current selected samplerate. */ uint64_t samplerate; + gboolean samplerate_sent; /* Number of channels. */ - unsigned int num_channels; + size_t num_channels; /* Column delimiter character(s). */ GString *delimiter; @@ -89,27 +144,21 @@ struct context { /* Comment prefix character(s). */ GString *comment; - /* Termination character(s) used in current stream. */ + /* 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. */ - unsigned int single_column; + /* Parameters how to process the columns. */ + size_t column_want_count; + struct column_details *column_details; - /* - * 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. - */ - unsigned int first_column; + /* Column number of the sample data in single column mode. */ + size_t single_column; - /* - * Column number of the first channel in multi column mode and position of - * the bit for the first channel in single column mode. - */ - unsigned int first_channel; + /* Column number of the first channel. */ + size_t first_column; /* Line number to start processing. */ size_t start_line; @@ -118,262 +167,340 @@ struct context { * Determines if the first line should be treated as header and used for * channel names in multi column mode. */ - gboolean header; + gboolean use_header; + gboolean header_seen; /* Format sample data is stored in single column mode. */ - int format; + enum single_col_format format; - /* Size of the sample buffer. */ - size_t sample_buffer_size; + size_t sample_unit_size; /**!< Byte count for a single sample. */ + uint8_t *sample_buffer; /**!< Buffer for a single sample. */ - /* Buffer to store sample data. */ - uint8_t *sample_buffer; + uint8_t *datafeed_buffer; /**!< Queue for datafeed submission. */ + size_t datafeed_buf_size; + size_t datafeed_buf_fill; /* Current line number. */ size_t line_number; + + /* List of previously created sigrok channels. */ + 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->num_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 %u 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->num_channels + 7) >> 3); + 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; - i = inc->first_channel; - - 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 %u 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 (!length) { - sr_err("Column %u 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->num_channels + 7) >> 3); - - /* Calculate the position of the first hexadecimal digit. */ - i = inc->first_channel / 4; - - for (j = 0; i < length && j < inc->num_channels; i++) { - c = str[length - i - 1]; - - if (!g_ascii_isxdigit(c)) { - sr_err("Invalid value '%s' in column %u in line %zu.", - str, inc->single_column, inc->line_number); - return SR_ERR; - } - - value = g_ascii_xdigit_value(c); +static int make_column_details_single(struct context *inc, + size_t col_nr, size_t bit_count, enum single_col_format format) +{ + struct column_details *details; - k = (inc->first_channel + j) % 4; + /* + * Need at least as many columns to also include the one with + * the single-column input data. + */ + inc->column_want_count = col_nr; - for (; j < inc->num_channels && k < 4; k++) { - if (value & (1 << k)) - inc->sample_buffer[j / 8] |= (1 << (j % 8)); + /* + * Allocate the columns' processing details. Columns are counted + * from 1 (user's perspective), array items from 0 (programmer's + * perspective). + */ + inc->column_details = g_malloc0_n(col_nr, sizeof(inc->column_details[0])); + details = &inc->column_details[col_nr - 1]; + details->col_nr = col_nr; - j++; - } - } + /* + * In single-column mode this single column will hold all bits + * of all logic channels, in the user specified number format. + */ + details->text_format = format; + details->channel_offset = 0; + details->channel_count = bit_count; return SR_OK; } -static int parse_octstr(const char *str, struct context *inc) +static int make_column_details_multi(struct context *inc, + size_t first_col, size_t last_col) { - gsize i, j, k, length; - uint8_t value; - char c; + struct column_details *details; + size_t col_nr; - length = strlen(str); + /* + * Need at least as many columns to also include the one with + * the last channel's data. + */ + inc->column_want_count = last_col; - if (!length) { - sr_err("Column %u in line %zu is empty.", inc->single_column, - inc->line_number); - return SR_ERR; + /* + * Allocate the columns' processing details. Columns are counted + * from 1, array items from 0. + * In multi-column mode each column will hold a single bit for + * the respective channel. + */ + inc->column_details = g_malloc0_n(last_col, sizeof(inc->column_details[0])); + for (col_nr = first_col; col_nr <= last_col; col_nr++) { + details = &inc->column_details[col_nr - 1]; + details->col_nr = col_nr; + details->text_format = FORMAT_BIN; + details->channel_offset = col_nr - first_col; + details->channel_count = 1; } - /* Clear buffer in order to set bits only. */ - memset(inc->sample_buffer, 0, (inc->num_channels + 7) >> 3); - - /* 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 %u in line %zu.", - str, inc->single_column, inc->line_number); - return SR_ERR; - } - - value = g_ascii_xdigit_value(c); - - k = (inc->first_channel + j) % 3; - - for (; j < inc->num_channels && k < 3; k++) { - if (value & (1 << k)) - inc->sample_buffer[j / 8] |= (1 << (j % 8)); - - j++; - } - } return SR_OK; } -static char **parse_line(char *buf, struct context *inc, int max_columns) +static const struct column_details *lookup_column_details(struct context *inc, size_t nr) { - 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_columns) { - if (n >= inc->first_column) { - column = g_strndup(remainder, str - remainder); - list = g_slist_prepend(list, g_strstrip(column)); - - max_columns--; - k++; - } - - remainder = str + inc->delimiter->len; - str = strstr(remainder, inc->delimiter->str); - n++; - } - - if (buf[0] && max_columns && 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))) + if (!inc || !inc->column_details) return NULL; - - columns[k--] = NULL; - - for (l = list; l; l = l->next) - columns[k--] = l->data; - - g_slist_free(list); - - return columns; + if (!nr || nr > inc->column_want_count) + return NULL; + return &inc->column_details[nr - 1]; } -static int parse_multi_columns(char **columns, struct context *inc) +/* + * Primitive operations for text input: Strip comments off text lines. + * Split text lines into columns. Process input text for individual + * columns. + */ + +static void strip_comment(char *buf, const GString *prefix) { - gsize i; + char *ptr; - /* Clear buffer in order to set bits only. */ - memset(inc->sample_buffer, 0, (inc->num_channels + 7) >> 3); + if (!prefix->len) + return; - for (i = 0; i < inc->num_channels; i++) { - if (columns[i][0] == '1') { - inc->sample_buffer[i / 8] |= (1 << (i % 8)); - } else if (!strlen(columns[i])) { - sr_err("Column %zu in line %zu is empty.", - inc->first_channel + i, inc->line_number); - return SR_ERR; - } else if (columns[i][0] != '0') { - sr_err("Invalid value '%s' in column %zu in line %zu.", - columns[i], inc->first_channel + i, - inc->line_number); - return SR_ERR; - } + if ((ptr = strstr(buf, prefix->str))) { + *ptr = '\0'; + g_strstrip(buf); } - - return SR_OK; } -static int parse_single_column(const char *column, struct context *inc) +/** + * @brief Splits a text line into a set of columns. + * + * @param[in] buf The input text line to split. + * @param[in] inc The input module's context. + * + * @returns An array of strings, representing the columns' text. + * + * This routine splits a text line on previously determined separators. + */ +static char **split_line(char *buf, 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; + return g_strsplit(buf, inc->delimiter->str, 0); } -static int send_samples(const struct sr_dev_inst *sdi, uint8_t *buffer, - gsize buffer_size, gsize count) +/** + * @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] col_nr The involved column number (1-based). + * + * @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_column(const char *column, struct context *inc, size_t col_nr) { - struct sr_datafeed_packet packet; - struct sr_datafeed_logic logic; - int res; - gsize i; + const struct column_details *col_det; + size_t length, ch_rem, ch_idx, ch_inc; + const char *rdptr; + char c; + gboolean valid; + const char *type_text; + uint8_t bits; + + /* See whether and how the columns needs to get processed. */ + col_det = lookup_column_details(inc, col_nr); + if (!col_det) { + sr_dbg("Column %zu in line %zu without processing details?", + col_nr, inc->line_number); + return SR_ERR; + } + if (col_det->text_format == FORMAT_NONE) + return SR_OK; - packet.type = SR_DF_LOGIC; - packet.payload = &logic; - logic.unitsize = buffer_size; - logic.length = buffer_size; - logic.data = buffer; + /* + * 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.", col_nr, + inc->line_number); + return SR_ERR; + } + rdptr = &column[length]; + ch_idx = col_det->channel_offset; + ch_rem = col_det->channel_count; - for (i = 0; i < count; i++) { - if ((res = sr_session_send(sdi, &packet)) != SR_OK) - return res; + /* + * 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 (inc->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[inc->format]; + sr_err("Invalid text '%s' in %s type column %zu in line %zu.", + column, type_text, col_nr, inc->line_number); + return SR_ERR; + } + /* Use the digit's bits for logic channels' data. */ + bits = g_ascii_xdigit_value(c); + switch (inc->format) { + case FORMAT_NONE: + /* ShouldNotHappen(TM), but silences compiler warning. */ + return SR_ERR; + 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; + } + 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; } @@ -382,19 +509,20 @@ static int init(struct sr_input *in, GHashTable *options) { struct context *inc; const char *s; + int ret; in->sdi = g_malloc0(sizeof(struct sr_dev_inst)); in->priv = inc = g_malloc0(sizeof(struct context)); - inc->single_column = g_variant_get_int32(g_hash_table_lookup(options, "single-column")); + inc->single_column = g_variant_get_uint32(g_hash_table_lookup(options, "single-column")); inc->multi_column_mode = inc->single_column == 0; - inc->num_channels = g_variant_get_int32(g_hash_table_lookup(options, "numchannels")); + inc->num_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."); + sr_err("Column delimiter cannot be empty."); return SR_ERR_ARG; } @@ -413,37 +541,109 @@ 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_int32(g_hash_table_lookup(options, "first-channel")); + inc->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_int32(g_hash_table_lookup(options, "startline")); + inc->start_line = g_variant_get_uint32(g_hash_table_lookup(options, "startline")); if (inc->start_line < 1) { sr_err("Invalid start line %zu.", inc->start_line); 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."); + /* + * Derive the set of columns to inspect and their respective + * formats from simple input specs. Remain close to the previous + * set of option keywords and their meaning. Exclusively support + * a single column with multiple bits in it, or an adjacent set + * of colums with one bit each. The latter may not know the total + * column count here (when the user omitted the spec), and will + * derive it from the first text line of the input file. + */ + if (inc->single_column && inc->num_channels) { + sr_dbg("DIAG Got single column (%zu) and channels (%zu).", + inc->single_column, inc->num_channels); + sr_dbg("DIAG -> column %zu, %zu bits in %s format.", + inc->single_column, inc->num_channels, + col_format_text[inc->format]); + ret = make_column_details_single(inc, + inc->single_column, inc->num_channels, inc->format); + if (ret != SR_OK) + return ret; + } else if (inc->multi_column_mode) { + sr_dbg("DIAG Got multi-column, first column %zu, count %zu.", + inc->first_column, inc->num_channels); + if (inc->num_channels) { + sr_dbg("DIAG -> columns %zu-%zu, 1 bit each.", + inc->first_column, + inc->first_column + inc->num_channels - 1); + ret = make_column_details_multi(inc, inc->first_column, + inc->first_column + inc->num_channels - 1); + if (ret != SR_OK) + return ret; + } else { + sr_dbg("DIAG -> incomplete spec, have to update later."); + } + } else { + sr_err("Unknown or unsupported combination of option values."); return SR_ERR_ARG; } return SR_OK; } +/* + * Check the channel list for consistency across file re-import. See + * the VCD input module for more details and motivation. + */ + +static void keep_header_for_reread(const struct sr_input *in) +{ + struct context *inc; + + inc = in->priv; + g_slist_free_full(inc->prev_sr_channels, sr_channel_free_cb); + inc->prev_sr_channels = in->sdi->channels; + in->sdi->channels = NULL; +} + +static int check_header_in_reread(const struct sr_input *in) +{ + struct context *inc; + + if (!in) + return FALSE; + inc = in->priv; + if (!inc) + return FALSE; + if (!inc->prev_sr_channels) + return TRUE; + + if (sr_channel_lists_differ(inc->prev_sr_channels, in->sdi->channels)) { + sr_err("Channel list change not supported for file re-read."); + return FALSE; + } + g_slist_free_full(in->sdi->channels, sr_channel_free_cb); + in->sdi->channels = inc->prev_sr_channels; + inc->prev_sr_channels = NULL; + + return TRUE; +} + +static const char *delim_set = "\r\n"; + static const char *get_line_termination(GString *buf) { const char *term; @@ -463,29 +663,32 @@ static int initial_parse(const struct sr_input *in, GString *buf) { struct context *inc; GString *channel_name; - unsigned int 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, **columns; + 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, "\r\n", 0); - for (l = 0; lines[l]; l++) { + lines = g_strsplit_set(buf->str, delim_set, 0); + for (line_idx = 0; (line = lines[line_idx]); line_idx++) { line_number++; if (inc->start_line > line_number) { - sr_spew("Line %zu skipped.", line_number); + sr_spew("Line %zu skipped (before start).", line_number); continue; } - if (lines[l][0] == '\0') { + if (line[0] == '\0') { sr_spew("Blank line %zu skipped.", line_number); continue; } - strip_comment(lines[l], inc->comment); - if (lines[l][0] == '\0') { + strip_comment(line, inc->comment); + if (line[0] == '\0') { sr_spew("Comment-only line %zu skipped.", line_number); continue; } @@ -493,70 +696,105 @@ 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. - */ - if (!(columns = parse_line(lines[l], 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 %u 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; } - - 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: %u.", - inc->num_channels); - } - - /* - * 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; + sr_dbg("DIAG Got %zu columns in text line: %s.", num_columns, line); + + /* Optionally update incomplete multi-column specs. */ + if (inc->multi_column_mode && !inc->num_channels) { + inc->num_channels = num_columns - inc->first_column + 1; + sr_dbg("DIAG -> multi-column update: columns %zu-%zu, 1 bit each.", + inc->first_column, + inc->first_column + inc->num_channels - 1); + ret = make_column_details_multi(inc, inc->first_column, + inc->first_column + inc->num_channels - 1); + if (ret != SR_OK) 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++) { - if (inc->header && inc->multi_column_mode && columns[i][0] != '\0') - g_string_assign(channel_name, columns[i]); - else - g_string_printf(channel_name, "%u", 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)) { + ret = SR_ERR_DATA; + goto out; + } /* - * Calculate the minimum buffer size to store the sample data of the - * channels. + * 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, the "sample buffer" will point + * to a location within that large buffer later. */ - inc->sample_buffer_size = (inc->num_channels + 7) >> 3; - inc->sample_buffer = g_malloc(inc->sample_buffer_size); + inc->sample_unit_size = (inc->num_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; out: if (columns) @@ -566,6 +804,27 @@ out: return ret; } +/* + * Gets called from initial_receive(), which runs until the end-of-line + * encoding of the input stream could get determined. Assumes that this + * routine receives enough buffered initial input data to either see the + * BOM when there is one, or that no BOM will follow when a text line + * termination sequence was seen. Silently drops the UTF-8 BOM sequence + * from the input buffer if one was seen. Does not care to protect + * against multiple execution or dropping the BOM multiple times -- + * there should be at most one in the input stream. + */ +static void initial_bom_check(const struct sr_input *in) +{ + static const char *utf8_bom = "\xef\xbb\xbf"; + + if (in->buf->len < strlen(utf8_bom)) + return; + if (strncmp(in->buf->str, utf8_bom, strlen(utf8_bom)) != 0) + return; + g_string_erase(in->buf, 0, strlen(utf8_bom)); +} + static int initial_receive(const struct sr_input *in) { struct context *inc; @@ -574,13 +833,17 @@ static int initial_receive(const struct sr_input *in) char *p; const char *termination; + initial_bom_check(in); + inc = in->priv; - if (!(termination = get_line_termination(in->buf))) + termination = get_line_termination(in->buf); + if (!termination) /* Don't have a full line yet. */ return SR_ERR_NA; - if (!(p = g_strrstr_len(in->buf->str, in->buf->len, termination))) + p = g_strrstr_len(in->buf->str, in->buf->len, termination); + if (!p) /* Don't have a full line yet. */ return SR_ERR_NA; len = p - in->buf->str - 1; @@ -599,113 +862,111 @@ static int initial_receive(const struct sr_input *in) return ret; } -static int process_buffer(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; - int max_columns, ret, l; - char *p, **lines, **columns; + size_t line_idx, col_idx, col_nr; + int ret; + char *p, **lines, *line, **columns, *column; inc = in->priv; if (!inc->started) { - std_session_send_df_header(in->sdi, LOG_PREFIX); - - 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); - } - + std_session_send_df_header(in->sdi); inc->started = TRUE; } - if (!(p = g_strrstr_len(in->buf->str, in->buf->len, inc->termination))) - /* Don't have a full line. */ - return SR_ERR; - - *p = '\0'; + /* + * Consider empty input non-fatal. Keep accumulating input until + * at least one full text line has become available. Grab the + * maximum amount of accumulated data that consists of full text + * lines, and process what has been received so far, leaving not + * yet complete lines for the next invocation. + * + * Enforce that all previously buffered data gets processed in + * the "EOF" condition. Do not insist in the presence of the + * termination sequence for the last line (may often be missing + * on Windows). A present termination sequence will just result + * in the "execution of an empty line", and does not harm. + */ + if (!in->buf->len) + return SR_OK; + if (is_eof) { + p = in->buf->str + in->buf->len; + } else { + p = g_strrstr_len(in->buf->str, in->buf->len, inc->termination); + if (!p) + return SR_ERR; + *p = '\0'; + p += strlen(inc->termination); + } g_strstrip(in->buf->str); - /* Limit the number of columns to parse. */ - if (inc->multi_column_mode) - max_columns = inc->num_channels; - else - max_columns = 1; - ret = SR_OK; - lines = g_strsplit_set(in->buf->str, "\r\n", 0); - for (l = 0; lines[l]; l++) { + lines = g_strsplit_set(in->buf->str, delim_set, 0); + for (line_idx = 0; (line = lines[line_idx]); line_idx++) { inc->line_number++; - if (lines[l][0] == '\0') { + if (line[0] == '\0') { sr_spew("Blank line %zu skipped.", inc->line_number); continue; } /* Remove trailing comment. */ - strip_comment(lines[l], inc->comment); - if (lines[l][0] == '\0') { + strip_comment(line, inc->comment); + if (line[0] == '\0') { sr_spew("Comment-only line %zu skipped.", inc->line_number); continue; } /* 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; } - if (!(columns = parse_line(lines[l], 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 %u in line %zu is out of bounds.", - inc->first_column, inc->line_number); - g_strfreev(columns); - 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); - return SR_ERR; + 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; + ret = parse_column(column, inc, col_nr); + if (ret != SR_OK) { + g_strfreev(columns); + g_strfreev(lines); + return SR_ERR; + } } - /* Send sample data to the session bus. */ - ret = send_samples(in->sdi, inc->sample_buffer, - inc->sample_buffer_size, 1); + /* 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); + g_strfreev(lines); return SR_ERR; } + g_strfreev(columns); } g_strfreev(lines); - g_string_erase(in->buf, 0, p - in->buf->str + 1); + g_string_erase(in->buf, 0, p - in->buf->str); return ret; } @@ -719,7 +980,8 @@ static int receive(struct sr_input *in, GString *buf) inc = in->priv; if (!inc->termination) { - if ((ret = initial_receive(in)) == SR_ERR_NA) + ret = initial_receive(in); + if (ret == SR_ERR_NA) /* Not enough data yet. */ return SR_OK; else if (ret != SR_OK) @@ -730,7 +992,7 @@ static int receive(struct sr_input *in, GString *buf) return SR_OK; } - ret = process_buffer(in); + ret = process_buffer(in, FALSE); return ret; } @@ -741,13 +1003,19 @@ static int end(struct sr_input *in) int ret; if (in->sdi_ready) - ret = process_buffer(in); + ret = process_buffer(in, TRUE); else ret = SR_OK; + if (ret != SR_OK) + return ret; + + ret = flush_logic_samples(in); + if (ret != SR_OK) + return ret; inc = in->priv; if (inc->started) - std_session_send_df_end(in->sdi, LOG_PREFIX); + std_session_send_df_end(in->sdi); return ret; } @@ -756,43 +1024,72 @@ static void cleanup(struct sr_input *in) { struct context *inc; + keep_header_for_reread(in); + inc = in->priv; - if (inc->delimiter) - g_string_free(inc->delimiter, TRUE); + g_free(inc->termination); + inc->termination = NULL; + g_free(inc->datafeed_buffer); + inc->datafeed_buffer = NULL; +} - if (inc->comment) - g_string_free(inc->comment, TRUE); +static int reset(struct sr_input *in) +{ + struct context *inc = in->priv; - g_free(inc->termination); - g_free(inc->sample_buffer); + cleanup(in); + inc->started = FALSE; + g_string_truncate(in->buf, 0); + + return SR_OK; } +enum option_index { + OPT_SINGLE_COL, + OPT_NUM_LOGIC, + OPT_DELIM, + OPT_FORMAT, + OPT_COMMENT, + OPT_RATE, + OPT_FIRST_COL, + OPT_HEADER, + OPT_START, + OPT_MAX, +}; + static struct sr_option options[] = { - { "single-column", "Single column", "Enable/specify single column", NULL, NULL }, - { "numchannels", "Max channels", "Number of channels", NULL, NULL }, - { "delimiter", "Delimiter", "Column delimiter", NULL, NULL }, - { "format", "Format", "Numeric format", NULL, NULL }, - { "comment", "Comment", "Comment prefix character", NULL, NULL }, - { "samplerate", "Samplerate", "Samplerate used during capture", NULL, NULL }, - { "first-channel", "First channel", "Column number of first channel", NULL, NULL }, - { "header", "Header", "Treat first line as header with channel names", NULL, NULL }, - { "startline", "Start line", "Line number at which to start processing samples", NULL, NULL }, - ALL_ZERO + [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_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, }; static const struct sr_option *get_options(void) { + GSList *l; + if (!options[0].def) { - options[0].def = g_variant_ref_sink(g_variant_new_int32(0)); - options[1].def = g_variant_ref_sink(g_variant_new_int32(0)); - options[2].def = g_variant_ref_sink(g_variant_new_string(",")); - options[3].def = g_variant_ref_sink(g_variant_new_string("bin")); - options[4].def = g_variant_ref_sink(g_variant_new_string(";")); - options[5].def = g_variant_ref_sink(g_variant_new_uint64(0)); - options[6].def = g_variant_ref_sink(g_variant_new_int32(0)); - options[7].def = g_variant_ref_sink(g_variant_new_boolean(FALSE)); - options[8].def = g_variant_ref_sink(g_variant_new_int32(1)); + 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; + l = g_slist_append(l, g_variant_ref_sink(g_variant_new_string("bin"))); + l = g_slist_append(l, g_variant_ref_sink(g_variant_new_string("hex"))); + l = g_slist_append(l, g_variant_ref_sink(g_variant_new_string("oct"))); + 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_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_uint32(1)); } return options; @@ -808,4 +1105,5 @@ SR_PRIV struct sr_input_module input_csv = { .receive = receive, .end = end, .cleanup = cleanup, + .reset = reset, };