1 /* PSPP - a program for statistical analysis.
2 Copyright (C) 1997-2004, 2006 Free Software Foundation, Inc.
4 This program is free software: you can redistribute it and/or modify
5 it under the terms of the GNU General Public License as published by
6 the Free Software Foundation, either version 3 of the License, or
7 (at your option) any later version.
9 This program is distributed in the hope that it will be useful,
10 but WITHOUT ANY WARRANTY; without even the implied warranty of
11 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 GNU General Public License for more details.
14 You should have received a copy of the GNU General Public License
15 along with this program. If not, see <http://www.gnu.org/licenses/>. */
19 #include <language/data-io/data-reader.h>
27 #include <data/casereader.h>
28 #include <data/file-handle-def.h>
29 #include <data/file-name.h>
30 #include <data/procedure.h>
31 #include <language/command.h>
32 #include <language/data-io/file-handle.h>
33 #include <language/lexer/lexer.h>
34 #include <language/prompt.h>
35 #include <libpspp/assertion.h>
36 #include <libpspp/message.h>
37 #include <libpspp/str.h>
43 #define _(msgid) gettext (msgid)
45 /* Flags for DFM readers. */
48 DFM_ADVANCE = 002, /* Read next line on dfm_get_record() call? */
49 DFM_SAW_BEGIN_DATA = 004, /* For inline_file only, whether we've
50 already read a BEGIN DATA line. */
51 DFM_TABS_EXPANDED = 010, /* Tabs have been expanded. */
54 /* Data file reader. */
57 struct file_handle *fh; /* File handle. */
58 struct fh_lock *lock; /* Mutual exclusion lock for file. */
59 struct msg_locator where; /* Current location in data file. */
60 struct string line; /* Current line. */
61 struct string scratch; /* Extra line buffer. */
62 enum dfm_reader_flags flags; /* Zero or more of DFM_*. */
63 FILE *file; /* Associated file. */
64 size_t pos; /* Offset in line of current character. */
65 unsigned eof_cnt; /* # of attempts to advance past EOF. */
66 struct lexer *lexer; /* The lexer reading the file */
69 /* Closes reader R opened by dfm_open_reader(). */
71 dfm_close_reader (struct dfm_reader *r)
76 if (fh_unlock (r->lock))
78 /* File is still locked by another client. */
82 /* This was the last client, so close the underlying file. */
83 if (fh_get_referent (r->fh) != FH_REF_INLINE)
84 fn_close (fh_get_file_name (r->fh), r->file);
87 /* Skip any remaining data on the inline file. */
88 if (r->flags & DFM_SAW_BEGIN_DATA)
90 dfm_reread_record (r, 0);
92 dfm_forward_record (r);
97 ds_destroy (&r->line);
98 ds_destroy (&r->scratch);
102 /* Opens the file designated by file handle FH for reading as a
103 data file. Providing fh_inline_file() for FH designates the
104 "inline file", that is, data included inline in the command
105 file between BEGIN FILE and END FILE. Returns a reader if
106 successful, or a null pointer otherwise. */
108 dfm_open_reader (struct file_handle *fh, struct lexer *lexer)
110 struct dfm_reader *r;
111 struct fh_lock *lock;
113 lock = fh_lock (fh, FH_REF_FILE | FH_REF_INLINE, "data file",
118 r = fh_lock_get_aux (lock);
122 r = xmalloc (sizeof *r);
126 ds_init_empty (&r->line);
127 ds_init_empty (&r->scratch);
128 r->flags = DFM_ADVANCE;
130 if (fh_get_referent (fh) != FH_REF_INLINE)
132 r->where.file_name = fh_get_file_name (fh);
133 r->where.line_number = 0;
134 r->file = fn_open (fh_get_file_name (fh), "rb");
137 msg (ME, _("Could not open \"%s\" for reading as a data file: %s."),
138 fh_get_file_name (r->fh), strerror (errno));
145 fh_lock_set_aux (lock, r);
150 /* Returns true if an I/O error occurred on READER, false otherwise. */
152 dfm_reader_error (const struct dfm_reader *r)
154 return fh_get_referent (r->fh) == FH_REF_FILE && ferror (r->file);
157 /* Reads a record from the inline file into R.
158 Returns true if successful, false on failure. */
160 read_inline_record (struct dfm_reader *r)
162 if ((r->flags & DFM_SAW_BEGIN_DATA) == 0)
164 r->flags |= DFM_SAW_BEGIN_DATA;
166 while (lex_token (r->lexer) == '.')
168 if (!lex_force_match_id (r->lexer, "BEGIN") || !lex_force_match_id (r->lexer, "DATA"))
170 prompt_set_style (PROMPT_DATA);
173 if (!lex_get_line_raw (r->lexer))
175 msg (SE, _("Unexpected end-of-file while reading data in BEGIN "
176 "DATA. This probably indicates "
177 "a missing or misformatted END DATA command. "
178 "END DATA must appear by itself on a single line "
179 "with exactly one space between words."));
183 if (ds_length (lex_entire_line_ds (r->lexer) ) >= 8
184 && !strncasecmp (lex_entire_line (r->lexer), "end data", 8))
186 lex_discard_line (r->lexer);
190 ds_assign_string (&r->line, lex_entire_line_ds (r->lexer) );
195 /* Reads a record from a disk file into R.
196 Returns true if successful, false on failure. */
198 read_file_record (struct dfm_reader *r)
200 assert (r->fh != fh_inline_file ());
202 if (fh_get_mode (r->fh) == FH_MODE_TEXT)
204 if (!ds_read_line (&r->line, r->file))
206 if (ferror (r->file))
207 msg (ME, _("Error reading file %s: %s."),
208 fh_get_name (r->fh), strerror (errno));
211 ds_chomp (&r->line, '\n');
213 else if (fh_get_mode (r->fh) == FH_MODE_BINARY)
215 size_t record_width = fh_get_record_width (r->fh);
216 size_t amt = ds_read_stream (&r->line, 1, record_width, r->file);
217 if (record_width != amt)
219 if (ferror (r->file))
220 msg (ME, _("Error reading file %s: %s."),
221 fh_get_name (r->fh), strerror (errno));
223 msg (ME, _("%s: Partial record at end of file."),
224 fh_get_name (r->fh));
232 r->where.line_number++;
237 /* Reads a record from R, setting the current position to the
238 start of the line. If an error occurs or end-of-file is
239 encountered, the current line is set to null. */
241 read_record (struct dfm_reader *r)
243 return (fh_get_referent (r->fh) == FH_REF_FILE
244 ? read_file_record (r)
245 : read_inline_record (r));
248 /* Returns the number of attempts, thus far, to advance past
249 end-of-file in reader R. Reads forward in HANDLE's file, if
250 necessary, to find out.
252 Normally, the user stops attempting to read from the file the
253 first time EOF is reached (a return value of 1). If the user
254 tries to read past EOF again (a return value of 2 or more),
255 an error message is issued, and the caller should more
256 forcibly abort to avoid an infinite loop. */
258 dfm_eof (struct dfm_reader *r)
260 if (r->flags & DFM_ADVANCE)
262 r->flags &= ~DFM_ADVANCE;
264 if (r->eof_cnt == 0 && read_record (r) )
273 if (r->fh != fh_inline_file ())
274 msg (ME, _("Attempt to read beyond end-of-file on file %s."),
275 fh_get_name (r->fh));
277 msg (ME, _("Attempt to read beyond END DATA."));
284 /* Returns the current record in the file corresponding to
285 HANDLE. Aborts if reading from the file is necessary or at
286 end of file, so call dfm_eof() first. */
288 dfm_get_record (struct dfm_reader *r)
290 assert ((r->flags & DFM_ADVANCE) == 0);
291 assert (r->eof_cnt == 0);
293 return ds_substr (&r->line, r->pos, SIZE_MAX);
296 /* Expands tabs in the current line into the equivalent number of
297 spaces, if appropriate for this kind of file. Aborts if
298 reading from the file is necessary or at end of file, so call
301 dfm_expand_tabs (struct dfm_reader *r)
303 size_t ofs, new_pos, tab_width;
305 assert ((r->flags & DFM_ADVANCE) == 0);
306 assert (r->eof_cnt == 0);
308 if (r->flags & DFM_TABS_EXPANDED)
310 r->flags |= DFM_TABS_EXPANDED;
312 if (r->fh != fh_inline_file ()
313 && (fh_get_mode (r->fh) == FH_MODE_BINARY
314 || fh_get_tab_width (r->fh) == 0
315 || ds_find_char (&r->line, '\t') == SIZE_MAX))
318 /* Expand tabs from r->line into r->scratch, and figure out
319 new value for r->pos. */
320 tab_width = fh_get_tab_width (r->fh);
321 ds_clear (&r->scratch);
323 for (ofs = 0; ofs < ds_length (&r->line); ofs++)
328 new_pos = ds_length (&r->scratch);
330 c = ds_data (&r->line)[ofs];
332 ds_put_char (&r->scratch, c);
336 ds_put_char (&r->scratch, ' ');
337 while (ds_length (&r->scratch) % tab_width != 0);
340 if (new_pos == SIZE_MAX)
342 /* Maintain the same relationship between position and line
343 length that we had before. DATA LIST uses a
344 beyond-the-end position to deal with an empty field at
345 the end of the line. */
346 assert (r->pos >= ds_length (&r->line));
347 new_pos = (r->pos - ds_length (&r->line)) + ds_length (&r->scratch);
350 /* Swap r->line and r->scratch and set new r->pos. */
351 ds_swap (&r->line, &r->scratch);
355 /* Causes dfm_get_record() or dfm_get_whole_record() to read in
356 the next record the next time it is executed on file
359 dfm_forward_record (struct dfm_reader *r)
361 r->flags |= DFM_ADVANCE;
364 /* Cancels the effect of any previous dfm_fwd_record() executed
365 on file HANDLE. Sets the current line to begin in the 1-based
368 dfm_reread_record (struct dfm_reader *r, size_t column)
370 r->flags &= ~DFM_ADVANCE;
371 r->pos = MAX (column, 1) - 1;
374 /* Sets the current line to begin COLUMNS characters following
375 the current start. */
377 dfm_forward_columns (struct dfm_reader *r, size_t columns)
379 dfm_reread_record (r, (r->pos + 1) + columns);
382 /* Returns the 1-based column to which the line pointer in HANDLE
383 is set. Unless dfm_reread_record() or dfm_forward_columns()
384 have been called, this is 1. */
386 dfm_column_start (const struct dfm_reader *r)
391 /* Returns the number of columns we are currently beyond the end
392 of the line. At or before end-of-line, this is 0; one column
393 after end-of-line, this is 1; and so on. */
395 dfm_columns_past_end (const struct dfm_reader *r)
397 return r->pos < ds_length (&r->line) ? 0 : ds_length (&r->line) - r->pos;
400 /* Returns the 1-based column within the current line that P
403 dfm_get_column (const struct dfm_reader *r, const char *p)
405 return ds_pointer_to_position (&r->line, p) + 1;
408 /* Pushes the file name and line number on the fn/ln stack. */
410 dfm_push (struct dfm_reader *r)
412 if (r->fh != fh_inline_file ())
413 msg_push_msg_locator (&r->where);
416 /* Pops the file name and line number from the fn/ln stack. */
418 dfm_pop (struct dfm_reader *r)
420 if (r->fh != fh_inline_file ())
421 msg_pop_msg_locator (&r->where);
424 /* BEGIN DATA...END DATA procedure. */
426 /* Perform BEGIN DATA...END DATA as a procedure in itself. */
428 cmd_begin_data (struct lexer *lexer, struct dataset *ds)
430 struct dfm_reader *r;
433 if (!fh_is_locked (fh_inline_file (), FH_ACC_READ))
435 msg (SE, _("This command is not valid here since the current "
436 "input program does not access the inline file."));
437 return CMD_CASCADING_FAILURE;
440 /* Open inline file. */
441 r = dfm_open_reader (fh_inline_file (), lexer);
442 r->flags |= DFM_SAW_BEGIN_DATA;
444 /* Input procedure reads from inline file. */
445 prompt_set_style (PROMPT_DATA);
446 casereader_destroy (proc_open (ds));
447 ok = proc_commit (ds);
448 dfm_close_reader (r);
450 return ok ? CMD_SUCCESS : CMD_CASCADING_FAILURE;