| 1 |
<?php |
| 2 |
/** |
| 3 |
* CSV Parsing class for TablePress, used for import of CSV files |
| 4 |
* |
| 5 |
* @package TablePress |
| 6 |
* @subpackage Import |
| 7 |
* @author Tobias Bäthge |
| 8 |
* @since 1.0.0 |
| 9 |
*/ |
| 10 |
|
| 11 |
declare(strict_types=1); |
| 12 |
|
| 13 |
// Prohibit direct script loading. |
| 14 |
defined( 'ABSPATH' ) || die( 'No direct script access allowed!' ); |
| 15 |
|
| 16 |
/** |
| 17 |
* CSV Parsing class |
| 18 |
* |
| 19 |
* @package TablePress |
| 20 |
* @subpackage Import |
| 21 |
* @author Tobias Bäthge |
| 22 |
* @since 1.0.0 |
| 23 |
*/ |
| 24 |
class CSV_Parser { |
| 25 |
|
| 26 |
/** |
| 27 |
* The used character for the enclosure of a cell. Defaults to quotation mark ". |
| 28 |
* |
| 29 |
* @since 1.0.0 |
| 30 |
*/ |
| 31 |
protected string $enclosure = '"'; |
| 32 |
|
| 33 |
/** |
| 34 |
* Number of rows to analyze when attempting to auto-detect the CSV delimiter. |
| 35 |
* |
| 36 |
* @since 1.0.0 |
| 37 |
*/ |
| 38 |
protected int $delimiter_search_max_lines = 15; |
| 39 |
|
| 40 |
/** |
| 41 |
* Characters to ignore when attempting to auto-detect delimiter. |
| 42 |
* |
| 43 |
* @since 1.0.0 |
| 44 |
*/ |
| 45 |
protected string $non_delimiter_chars = "a-zA-Z0-9\n\r"; |
| 46 |
|
| 47 |
/** |
| 48 |
* The preferred delimiter characters, only used when all filtering method return multiple possible delimiters (happens very rarely). |
| 49 |
* There must not be more than 9 characters in the preferred delimiter character list, see `_check_delimiter_count()`. |
| 50 |
* |
| 51 |
* @since 1.0.0 |
| 52 |
*/ |
| 53 |
protected string $preferred_delimiter_chars = ";,\t"; |
| 54 |
|
| 55 |
/** |
| 56 |
* The CSV data string that shall be parsed to an array. |
| 57 |
* |
| 58 |
* @since 1.0.0 |
| 59 |
*/ |
| 60 |
protected string $import_data = ''; |
| 61 |
|
| 62 |
/** |
| 63 |
* The error state while parsing input data. |
| 64 |
* |
| 65 |
* 0 = No errors found. Everything should be fine. |
| 66 |
* 1 = A hopefully correctable syntax error was found. |
| 67 |
* 2 = The enclosure character was found in a non-enclosed field. This means the file is either corrupt, |
| 68 |
* or does not follow the common CSV standard. Please validate the parsed data manually. |
| 69 |
* |
| 70 |
* @since 1.0.0 |
| 71 |
*/ |
| 72 |
public int $error = 0; |
| 73 |
|
| 74 |
/** |
| 75 |
* Detailed error information. |
| 76 |
* |
| 77 |
* @since 1.0.0 |
| 78 |
* @var array<string, array<string, int|string>> |
| 79 |
*/ |
| 80 |
public array $error_info = array(); |
| 81 |
|
| 82 |
/** |
| 83 |
* Class Constructor. |
| 84 |
* |
| 85 |
* @since 1.0.0 |
| 86 |
*/ |
| 87 |
public function __construct() { |
| 88 |
// Unused. |
| 89 |
} |
| 90 |
|
| 91 |
/** |
| 92 |
* Load CSV data that shall be parsed. |
| 93 |
* |
| 94 |
* @since 1.0.0 |
| 95 |
* |
| 96 |
* @param string $data Data to be parsed. |
| 97 |
*/ |
| 98 |
public function load_data( string $data ): void { |
| 99 |
// Check for mandatory trailing line break. |
| 100 |
if ( ! str_ends_with( $data, "\n" ) ) { |
| 101 |
$data .= "\n"; |
| 102 |
} |
| 103 |
$this->import_data = $data; |
| 104 |
} |
| 105 |
|
| 106 |
/** |
| 107 |
* Detect the CSV delimiter, by analyzing some rows to determine the most probable delimiter character. |
| 108 |
* |
| 109 |
* @since 1.0.0 |
| 110 |
* |
| 111 |
* @return string Most probable delimiter character. |
| 112 |
*/ |
| 113 |
public function find_delimiter(): string { |
| 114 |
$data = &$this->import_data; |
| 115 |
|
| 116 |
$delimiter_count = array(); |
| 117 |
$enclosed = false; |
| 118 |
$current_line = 0; |
| 119 |
|
| 120 |
// Walk through each character in the CSV string (up to $this->delimiter_search_max_lines) and search potential delimiter characters. |
| 121 |
$data_length = strlen( $data ); |
| 122 |
for ( $i = 0; $i < $data_length; $i++ ) { |
| 123 |
$prev_char = ( $i - 1 >= 0 ) ? $data[ $i - 1 ] : ''; |
| 124 |
$curr_char = $data[ $i ]; |
| 125 |
$next_char = ( $i + 1 < $data_length ) ? $data[ $i + 1 ] : ''; |
| 126 |
|
| 127 |
if ( $curr_char === $this->enclosure ) { |
| 128 |
// Open and closing quotes. |
| 129 |
if ( ! $enclosed || $next_char !== $this->enclosure ) { |
| 130 |
$enclosed = ! $enclosed; // Flip bool. |
| 131 |
} elseif ( $enclosed ) { |
| 132 |
++$i; // Skip next character. |
| 133 |
} |
| 134 |
} elseif ( ( ( "\n" === $curr_char && "\r" !== $prev_char ) || "\r" === $curr_char ) && ! $enclosed ) { |
| 135 |
// Reached end of a line. |
| 136 |
++$current_line; |
| 137 |
if ( $current_line >= $this->delimiter_search_max_lines ) { |
| 138 |
break; |
| 139 |
} |
| 140 |
} elseif ( ! $enclosed ) { |
| 141 |
// At this point, $curr_char seems to be used as a delimiter, as it is not enclosed. |
| 142 |
// Count $curr_char if it is not in the $this->non_delimiter_chars list. |
| 143 |
if ( 0 === preg_match( '#[' . $this->non_delimiter_chars . ']#i', $curr_char ) ) { |
| 144 |
if ( ! isset( $delimiter_count[ $curr_char ][ $current_line ] ) ) { |
| 145 |
$delimiter_count[ $curr_char ][ $current_line ] = 0; // Initialize empty. |
| 146 |
} |
| 147 |
++$delimiter_count[ $curr_char ][ $current_line ]; |
| 148 |
} |
| 149 |
} |
| 150 |
} |
| 151 |
|
| 152 |
// Find most probable delimiter, by sorting their counts. |
| 153 |
$potential_delimiters = array(); |
| 154 |
foreach ( $delimiter_count as $char => $line_counts ) { |
| 155 |
$is_possible_delimiter = $this->_check_delimiter_count( $char, $line_counts, $current_line ); |
| 156 |
if ( false !== $is_possible_delimiter ) { |
| 157 |
$potential_delimiters[ $is_possible_delimiter ] = $char; |
| 158 |
} |
| 159 |
} |
| 160 |
ksort( $potential_delimiters ); |
| 161 |
|
| 162 |
// If no valid delimiter was found, use the character that was found in most rows. |
| 163 |
if ( empty( $potential_delimiters ) ) { |
| 164 |
$delimiter_counts = array_map( 'count', $delimiter_count ); |
| 165 |
arsort( $delimiter_counts, SORT_NUMERIC ); |
| 166 |
$potential_delimiters = array_keys( $delimiter_counts ); |
| 167 |
} |
| 168 |
|
| 169 |
// If still no delimiter was found, fall back to a comma. |
| 170 |
if ( empty( $potential_delimiters ) ) { |
| 171 |
$potential_delimiters = array( ',' ); |
| 172 |
} |
| 173 |
|
| 174 |
// Return first array element, as that has the highest count. |
| 175 |
return array_shift( $potential_delimiters ); |
| 176 |
} |
| 177 |
|
| 178 |
/** |
| 179 |
* Check if passed character can be a delimiter, by checking counts in each line. |
| 180 |
* |
| 181 |
* @since 1.0.0 |
| 182 |
* |
| 183 |
* @param string $char Character to check. |
| 184 |
* @param int[] $line_counts Counts for the characters in the lines. |
| 185 |
* @param int $number_lines Number of lines. |
| 186 |
* @return bool|string False if delimiter is not possible, string to be used as a sort key if character could be a delimiter. |
| 187 |
*/ |
| 188 |
protected function _check_delimiter_count( string $char, array $line_counts, int $number_lines ) /* : bool|string */ { |
| 189 |
// Was the potential delimiter found in every line? |
| 190 |
if ( count( $line_counts ) !== $number_lines ) { |
| 191 |
return false; |
| 192 |
} |
| 193 |
|
| 194 |
// Check if the count in every line is the same (or one higher for an "almost"). |
| 195 |
$first = null; |
| 196 |
$equal = null; |
| 197 |
$almost = false; |
| 198 |
foreach ( $line_counts as $count ) { |
| 199 |
if ( is_null( $first ) ) { |
| 200 |
$first = $count; |
| 201 |
} elseif ( $count === $first && false !== $equal ) { |
| 202 |
$equal = true; |
| 203 |
} elseif ( $count === $first + 1 && false !== $equal ) { |
| 204 |
$equal = true; |
| 205 |
$almost = true; |
| 206 |
} else { |
| 207 |
$equal = false; |
| 208 |
} |
| 209 |
} |
| 210 |
// Check equality only if there's more than one line. |
| 211 |
if ( $number_lines > 1 && ! $equal ) { |
| 212 |
return false; |
| 213 |
} |
| 214 |
|
| 215 |
// At this point, count is equal in all lines, so determine a string to sort priority. |
| 216 |
$match = ( $almost ) ? 2 : 1; |
| 217 |
// There must not be more than 9 characters in the preferred delimiter character list. |
| 218 |
$pref = strpos( $this->preferred_delimiter_chars, $char ); |
| 219 |
if ( false === $pref ) { |
| 220 |
$pref = 9; |
| 221 |
} |
| 222 |
return $pref . $match . '.' . ( 99999 - $first ); |
| 223 |
} |
| 224 |
|
| 225 |
/** |
| 226 |
* Parse CSV string into a two-dimensional array. |
| 227 |
* |
| 228 |
* @since 1.0.0 |
| 229 |
* |
| 230 |
* @param string $delimiter Delimiter character for the CSV parsing. |
| 231 |
* @return array<int, array<int, string>> Two-dimensional array with the data from the CSV string. |
| 232 |
*/ |
| 233 |
public function parse( string $delimiter ): array { |
| 234 |
$data = &$this->import_data; |
| 235 |
|
| 236 |
// Filter delimiter from the list, if it is a whitespace character. |
| 237 |
$white_spaces = str_replace( $delimiter, '', " \t\x0B\0" ); |
| 238 |
|
| 239 |
$rows = array(); // Complete rows. |
| 240 |
$row = array(); // Row that is currently built. |
| 241 |
$column = 0; // Current column index. |
| 242 |
$cell_content = ''; // Content of the currently processed cell. |
| 243 |
$enclosed = false; |
| 244 |
$was_enclosed = false; // To determine if the cell content will be trimmed of whitespace (only for enclosed cells). |
| 245 |
|
| 246 |
// Walk through each character in the CSV string. |
| 247 |
$data_length = strlen( $data ); |
| 248 |
for ( $i = 0; $i < $data_length; $i++ ) { |
| 249 |
$curr_char = $data[ $i ]; |
| 250 |
$next_char = ( $i + 1 < $data_length ) ? $data[ $i + 1 ] : ''; |
| 251 |
|
| 252 |
if ( $curr_char === $this->enclosure ) { |
| 253 |
// Open/close quotes, and inline quotes. |
| 254 |
if ( ! $enclosed ) { |
| 255 |
if ( '' === ltrim( $cell_content, $white_spaces ) ) { |
| 256 |
$enclosed = true; |
| 257 |
$was_enclosed = true; |
| 258 |
} else { |
| 259 |
$this->error = 2; |
| 260 |
$error_line = count( $rows ) + 1; |
| 261 |
$error_column = $column + 1; |
| 262 |
if ( ! isset( $this->error_info[ "{$error_line}-{$error_column}" ] ) ) { |
| 263 |
$this->error_info[ "{$error_line}-{$error_column}" ] = array( |
| 264 |
'type' => 2, |
| 265 |
'info' => "Syntax error found in line {$error_line}. Non-enclosed fields can not contain double-quotes.", |
| 266 |
'line' => $error_line, |
| 267 |
'column' => $error_column, |
| 268 |
); |
| 269 |
} |
| 270 |
$cell_content .= $curr_char; |
| 271 |
} |
| 272 |
} elseif ( $next_char === $this->enclosure ) { |
| 273 |
// Enclosure character within enclosed cell (" encoded as ""). |
| 274 |
$cell_content .= $curr_char; |
| 275 |
++$i; // Skip next character. |
| 276 |
} elseif ( $next_char !== $delimiter && "\r" !== $next_char && "\n" !== $next_char ) { |
| 277 |
// for-loop (instead of while-loop) that skips whitespace. |
| 278 |
for ( $x = ( $i + 1 ); isset( $data[ $x ] ) && '' === ltrim( $data[ $x ], $white_spaces ); $x++ ) { // phpcs:ignore Generic.CodeAnalysis.ForLoopWithTestFunctionCall.NotAllowed,Generic.CodeAnalysis.EmptyStatement.DetectedFor |
| 279 |
// Action is in iterator check. |
| 280 |
} |
| 281 |
if ( $data[ $x ] === $delimiter ) { |
| 282 |
$enclosed = false; |
| 283 |
$i = $x; |
| 284 |
} else { |
| 285 |
if ( $this->error < 1 ) { |
| 286 |
$this->error = 1; |
| 287 |
} |
| 288 |
$error_line = count( $rows ) + 1; |
| 289 |
$error_column = $column + 1; |
| 290 |
if ( ! isset( $this->error_info[ "{$error_line}-{$error_column}" ] ) ) { |
| 291 |
$this->error_info[ "{$error_line}-{$error_column}" ] = array( |
| 292 |
'type' => 1, |
| 293 |
'info' => "Syntax error found in line {$error_line}. A single double-quote was found within an enclosed string. Enclosed double-quotes must be escaped with a second double-quote.", |
| 294 |
'line' => $error_line, |
| 295 |
'column' => $error_column, |
| 296 |
); |
| 297 |
} |
| 298 |
$cell_content .= $curr_char; |
| 299 |
$enclosed = false; |
| 300 |
} |
| 301 |
} else { |
| 302 |
// The " was the closing one for the cell. |
| 303 |
$enclosed = false; |
| 304 |
} |
| 305 |
} elseif ( ( $curr_char === $delimiter || "\n" === $curr_char || "\r" === $curr_char ) && ! $enclosed ) { |
| 306 |
// End of cell (by $delimiter), or end of line (by line break, and not enclosed!). |
| 307 |
|
| 308 |
$row[ $column ] = ( $was_enclosed ) ? $cell_content : trim( $cell_content ); |
| 309 |
$cell_content = ''; |
| 310 |
$was_enclosed = false; |
| 311 |
++$column; |
| 312 |
|
| 313 |
// End of line. |
| 314 |
if ( "\n" === $curr_char || "\r" === $curr_char ) { |
| 315 |
// Append completed row. |
| 316 |
$rows[] = $row; |
| 317 |
$row = array(); |
| 318 |
$column = 0; |
| 319 |
if ( "\r" === $curr_char && "\n" === $next_char ) { |
| 320 |
// Skip next character in \r\n line breaks. |
| 321 |
++$i; |
| 322 |
} |
| 323 |
} |
| 324 |
} else { |
| 325 |
// Append character to current cell. |
| 326 |
$cell_content .= $curr_char; |
| 327 |
} |
| 328 |
} |
| 329 |
|
| 330 |
return $rows; |
| 331 |
} |
| 332 |
|
| 333 |
} // class CSV_Parser |
| 334 |
|
