1
by brian
clean slate |
1 |
/* Copyright (C) 2000-2006 MySQL AB
|
2 |
||
3 |
This program is free software; you can redistribute it and/or modify
|
|
4 |
it under the terms of the GNU General Public License as published by
|
|
5 |
the Free Software Foundation; version 2 of the License.
|
|
6 |
||
7 |
This program is distributed in the hope that it will be useful,
|
|
8 |
but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
9 |
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
10 |
GNU General Public License for more details.
|
|
11 |
||
12 |
You should have received a copy of the GNU General Public License
|
|
13 |
along with this program; if not, write to the Free Software
|
|
14 |
Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA */
|
|
15 |
||
16 |
||
17 |
/* Copy data from a textfile to table */
|
|
18 |
||
243.1.17
by Jay Pipes
FINAL PHASE removal of mysql_priv.h (Bye, bye my friend.) |
19 |
#include <drizzled/server_includes.h> |
1
by brian
clean slate |
20 |
#include "sql_repl.h" |
202.3.6
by Monty Taylor
First pass at gettexizing the error messages. |
21 |
#include <drizzled/drizzled_error_messages.h> |
1
by brian
clean slate |
22 |
|
23 |
||
24 |
class READ_INFO { |
|
25 |
File file; |
|
26 |
uchar *buffer, /* Buffer for read text */ |
|
27 |
*end_of_buff; /* Data in bufferts ends here */ |
|
28 |
uint buff_length, /* Length of buffert */ |
|
29 |
max_length; /* Max length of row */ |
|
30 |
char *field_term_ptr,*line_term_ptr,*line_start_ptr,*line_start_end; |
|
31 |
uint field_term_length,line_term_length,enclosed_length; |
|
32 |
int field_term_char,line_term_char,enclosed_char,escape_char; |
|
33 |
int *stack,*stack_pos; |
|
34 |
bool found_end_of_line,start_of_line,eof; |
|
35 |
bool need_end_io_cache; |
|
36 |
IO_CACHE cache; |
|
37 |
NET *io_net; |
|
38 |
||
39 |
public: |
|
40 |
bool error,line_cuted,found_null,enclosed; |
|
41 |
uchar *row_start, /* Found row starts here */ |
|
42 |
*row_end; /* Found row ends here */ |
|
264.2.6
by Andrey Hristov
Constify the usage of CHARSET_INFO almost to the last place in the code. |
43 |
const CHARSET_INFO *read_charset; |
1
by brian
clean slate |
44 |
|
264.2.6
by Andrey Hristov
Constify the usage of CHARSET_INFO almost to the last place in the code. |
45 |
READ_INFO(File file,uint tot_length, const CHARSET_INFO * const cs, |
1
by brian
clean slate |
46 |
String &field_term,String &line_start,String &line_term, |
47 |
String &enclosed,int escape,bool get_it_from_net, bool is_fifo); |
|
48 |
~READ_INFO(); |
|
49 |
int read_field(); |
|
50 |
int read_fixed_length(void); |
|
51 |
int next_line(void); |
|
52 |
char unescape(char chr); |
|
53 |
int terminator(char *ptr,uint length); |
|
54 |
bool find_start_of_fields(); |
|
55 |
||
56 |
/*
|
|
57 |
We need to force cache close before destructor is invoked to log
|
|
58 |
the last read block
|
|
59 |
*/
|
|
60 |
void end_io_cache() |
|
61 |
{
|
|
62 |
::end_io_cache(&cache); |
|
63 |
need_end_io_cache = 0; |
|
64 |
}
|
|
65 |
||
66 |
/*
|
|
67 |
Either this method, or we need to make cache public
|
|
68 |
Arg must be set from mysql_load() since constructor does not see
|
|
69 |
either the table or THD value
|
|
70 |
*/
|
|
71 |
void set_io_cache_arg(void* arg) { cache.arg = arg; } |
|
72 |
};
|
|
73 |
||
327.2.4
by Brian Aker
Refactoring table.h |
74 |
static int read_fixed_length(THD *thd, COPY_INFO &info, TableList *table_list, |
1
by brian
clean slate |
75 |
List<Item> &fields_vars, List<Item> &set_fields, |
76 |
List<Item> &set_values, READ_INFO &read_info, |
|
307
by Brian Aker
Minor cleanups around ulong in kernel. |
77 |
uint32_t skip_lines, |
1
by brian
clean slate |
78 |
bool ignore_check_option_errors); |
327.2.4
by Brian Aker
Refactoring table.h |
79 |
static int read_sep_field(THD *thd, COPY_INFO &info, TableList *table_list, |
1
by brian
clean slate |
80 |
List<Item> &fields_vars, List<Item> &set_fields, |
81 |
List<Item> &set_values, READ_INFO &read_info, |
|
307
by Brian Aker
Minor cleanups around ulong in kernel. |
82 |
String &enclosed, uint32_t skip_lines, |
1
by brian
clean slate |
83 |
bool ignore_check_option_errors); |
84 |
||
85 |
static bool write_execute_load_query_log_event(THD *thd, |
|
86 |
bool duplicates, bool ignore, |
|
87 |
bool transactional_table, |
|
88 |
THD::killed_state killed_status); |
|
89 |
||
90 |
/*
|
|
91 |
Execute LOAD DATA query
|
|
92 |
||
93 |
SYNOPSYS
|
|
94 |
mysql_load()
|
|
95 |
thd - current thread
|
|
96 |
ex - sql_exchange object representing source file and its parsing rules
|
|
97 |
table_list - list of tables to which we are loading data
|
|
98 |
fields_vars - list of fields and variables to which we read
|
|
99 |
data from file
|
|
100 |
set_fields - list of fields mentioned in set clause
|
|
101 |
set_values - expressions to assign to fields in previous list
|
|
102 |
handle_duplicates - indicates whenever we should emit error or
|
|
103 |
replace row if we will meet duplicates.
|
|
104 |
ignore - - indicates whenever we should ignore duplicates
|
|
105 |
read_file_from_client - is this LOAD DATA LOCAL ?
|
|
106 |
||
107 |
RETURN VALUES
|
|
51.1.53
by Jay Pipes
Removed/replaced DBUG symbols and standardized TRUE/FALSE |
108 |
true - error / false - success
|
1
by brian
clean slate |
109 |
*/
|
110 |
||
327.2.4
by Brian Aker
Refactoring table.h |
111 |
int mysql_load(THD *thd,sql_exchange *ex,TableList *table_list, |
1
by brian
clean slate |
112 |
List<Item> &fields_vars, List<Item> &set_fields, |
113 |
List<Item> &set_values, |
|
114 |
enum enum_duplicates handle_duplicates, bool ignore, |
|
115 |
bool read_file_from_client) |
|
116 |
{
|
|
117 |
char name[FN_REFLEN]; |
|
118 |
File file; |
|
327.1.5
by Brian Aker
Refactor around classes. TABLE_LIST has been factored out of table.h |
119 |
Table *table= NULL; |
1
by brian
clean slate |
120 |
int error; |
121 |
String *field_term=ex->field_term,*escaped=ex->escaped; |
|
122 |
String *enclosed=ex->enclosed; |
|
123 |
bool is_fifo=0; |
|
124 |
LOAD_FILE_INFO lf_info; |
|
125 |
char *db = table_list->db; // This is never null |
|
126 |
/*
|
|
127 |
If path for file is not defined, we will use the current database.
|
|
128 |
If this is not set, we will use the directory where the table to be
|
|
129 |
loaded is located
|
|
130 |
*/
|
|
131 |
char *tdb= thd->db ? thd->db : db; // Result is never null |
|
307
by Brian Aker
Minor cleanups around ulong in kernel. |
132 |
uint32_t skip_lines= ex->skip_lines; |
1
by brian
clean slate |
133 |
bool transactional_table; |
134 |
THD::killed_state killed_status= THD::NOT_KILLED; |
|
135 |
||
136 |
if (escaped->length() > 1 || enclosed->length() > 1) |
|
137 |
{
|
|
138 |
my_message(ER_WRONG_FIELD_TERMINATORS,ER(ER_WRONG_FIELD_TERMINATORS), |
|
139 |
MYF(0)); |
|
51.1.53
by Jay Pipes
Removed/replaced DBUG symbols and standardized TRUE/FALSE |
140 |
return(true); |
1
by brian
clean slate |
141 |
}
|
142 |
if (open_and_lock_tables(thd, table_list)) |
|
51.1.53
by Jay Pipes
Removed/replaced DBUG symbols and standardized TRUE/FALSE |
143 |
return(true); |
1
by brian
clean slate |
144 |
if (setup_tables_and_check_access(thd, &thd->lex->select_lex.context, |
145 |
&thd->lex->select_lex.top_join_list, |
|
146 |
table_list, |
|
147 |
&thd->lex->select_lex.leaf_tables, true)) |
|
51.1.1
by Jay Pipes
Merged PatG's removal of various DBUG stuff with still keeping DBUG_ASSERT calls since they seem to be breaking test runs |
148 |
return(-1); |
1
by brian
clean slate |
149 |
|
150 |
/*
|
|
151 |
Let us emit an error if we are loading data to table which is used
|
|
152 |
in subselect in SET clause like we do it for INSERT.
|
|
153 |
||
154 |
The main thing to fix to remove this restriction is to ensure that the
|
|
155 |
table is marked to be 'used for insert' in which case we should never
|
|
156 |
mark this table as 'const table' (ie, one that has only one row).
|
|
157 |
*/
|
|
158 |
if (unique_table(thd, table_list, table_list->next_global, 0)) |
|
159 |
{
|
|
160 |
my_error(ER_UPDATE_TABLE_USED, MYF(0), table_list->table_name); |
|
51.1.53
by Jay Pipes
Removed/replaced DBUG symbols and standardized TRUE/FALSE |
161 |
return(true); |
1
by brian
clean slate |
162 |
}
|
163 |
||
164 |
table= table_list->table; |
|
165 |
transactional_table= table->file->has_transactions(); |
|
166 |
||
167 |
if (!fields_vars.elements) |
|
168 |
{
|
|
169 |
Field **field; |
|
170 |
for (field=table->field; *field ; field++) |
|
171 |
fields_vars.push_back(new Item_field(*field)); |
|
172 |
bitmap_set_all(table->write_set); |
|
173 |
table->timestamp_field_type= TIMESTAMP_NO_AUTO_SET; |
|
174 |
/*
|
|
175 |
Let us also prepare SET clause, altough it is probably empty
|
|
176 |
in this case.
|
|
177 |
*/
|
|
178 |
if (setup_fields(thd, 0, set_fields, MARK_COLUMNS_WRITE, 0, 0) || |
|
179 |
setup_fields(thd, 0, set_values, MARK_COLUMNS_READ, 0, 0)) |
|
51.1.53
by Jay Pipes
Removed/replaced DBUG symbols and standardized TRUE/FALSE |
180 |
return(true); |
1
by brian
clean slate |
181 |
}
|
182 |
else
|
|
183 |
{ // Part field list |
|
184 |
/* TODO: use this conds for 'WITH CHECK OPTIONS' */
|
|
185 |
if (setup_fields(thd, 0, fields_vars, MARK_COLUMNS_WRITE, 0, 0) || |
|
186 |
setup_fields(thd, 0, set_fields, MARK_COLUMNS_WRITE, 0, 0) || |
|
187 |
check_that_all_fields_are_given_values(thd, table, table_list)) |
|
51.1.53
by Jay Pipes
Removed/replaced DBUG symbols and standardized TRUE/FALSE |
188 |
return(true); |
1
by brian
clean slate |
189 |
/*
|
190 |
Check whenever TIMESTAMP field with auto-set feature specified
|
|
191 |
explicitly.
|
|
192 |
*/
|
|
193 |
if (table->timestamp_field) |
|
194 |
{
|
|
195 |
if (bitmap_is_set(table->write_set, |
|
196 |
table->timestamp_field->field_index)) |
|
197 |
table->timestamp_field_type= TIMESTAMP_NO_AUTO_SET; |
|
198 |
else
|
|
199 |
{
|
|
200 |
bitmap_set_bit(table->write_set, |
|
201 |
table->timestamp_field->field_index); |
|
202 |
}
|
|
203 |
}
|
|
204 |
/* Fix the expressions in SET clause */
|
|
205 |
if (setup_fields(thd, 0, set_values, MARK_COLUMNS_READ, 0, 0)) |
|
51.1.53
by Jay Pipes
Removed/replaced DBUG symbols and standardized TRUE/FALSE |
206 |
return(true); |
1
by brian
clean slate |
207 |
}
|
208 |
||
209 |
table->mark_columns_needed_for_insert(); |
|
210 |
||
211 |
uint tot_length=0; |
|
212 |
bool use_blobs= 0, use_vars= 0; |
|
213 |
List_iterator_fast<Item> it(fields_vars); |
|
214 |
Item *item; |
|
215 |
||
216 |
while ((item= it++)) |
|
217 |
{
|
|
218 |
Item *real_item= item->real_item(); |
|
219 |
||
220 |
if (real_item->type() == Item::FIELD_ITEM) |
|
221 |
{
|
|
222 |
Field *field= ((Item_field*)real_item)->field; |
|
223 |
if (field->flags & BLOB_FLAG) |
|
224 |
{
|
|
225 |
use_blobs= 1; |
|
226 |
tot_length+= 256; // Will be extended if needed |
|
227 |
}
|
|
228 |
else
|
|
229 |
tot_length+= field->field_length; |
|
230 |
}
|
|
231 |
else if (item->type() == Item::STRING_ITEM) |
|
232 |
use_vars= 1; |
|
233 |
}
|
|
234 |
if (use_blobs && !ex->line_term->length() && !field_term->length()) |
|
235 |
{
|
|
236 |
my_message(ER_BLOBS_AND_NO_TERMINATED,ER(ER_BLOBS_AND_NO_TERMINATED), |
|
237 |
MYF(0)); |
|
51.1.53
by Jay Pipes
Removed/replaced DBUG symbols and standardized TRUE/FALSE |
238 |
return(true); |
1
by brian
clean slate |
239 |
}
|
240 |
if (use_vars && !field_term->length() && !enclosed->length()) |
|
241 |
{
|
|
242 |
my_error(ER_LOAD_FROM_FIXED_SIZE_ROWS_TO_VAR, MYF(0)); |
|
51.1.53
by Jay Pipes
Removed/replaced DBUG symbols and standardized TRUE/FALSE |
243 |
return(true); |
1
by brian
clean slate |
244 |
}
|
245 |
||
246 |
/* We can't give an error in the middle when using LOCAL files */
|
|
247 |
if (read_file_from_client && handle_duplicates == DUP_ERROR) |
|
248 |
ignore= 1; |
|
249 |
||
250 |
if (read_file_from_client) |
|
251 |
{
|
|
252 |
(void)net_request_file(&thd->net,ex->file_name); |
|
253 |
file = -1; |
|
254 |
}
|
|
255 |
else
|
|
256 |
{
|
|
257 |
#ifdef DONT_ALLOW_FULL_LOAD_DATA_PATHS
|
|
258 |
ex->file_name+=dirname_length(ex->file_name); |
|
259 |
#endif
|
|
260 |
if (!dirname_length(ex->file_name)) |
|
261 |
{
|
|
262 |
strxnmov(name, FN_REFLEN-1, mysql_real_data_home, tdb, NullS); |
|
263 |
(void) fn_format(name, ex->file_name, name, "", |
|
264 |
MY_RELATIVE_PATH | MY_UNPACK_FILENAME); |
|
265 |
}
|
|
266 |
else
|
|
267 |
{
|
|
268 |
(void) fn_format(name, ex->file_name, mysql_real_data_home, "", |
|
269 |
MY_RELATIVE_PATH | MY_UNPACK_FILENAME); |
|
270 |
||
271 |
if (opt_secure_file_priv && |
|
272 |
strncmp(opt_secure_file_priv, name, strlen(opt_secure_file_priv))) |
|
273 |
{
|
|
274 |
/* Read only allowed from within dir specified by secure_file_priv */
|
|
275 |
my_error(ER_OPTION_PREVENTS_STATEMENT, MYF(0), "--secure-file-priv"); |
|
51.1.53
by Jay Pipes
Removed/replaced DBUG symbols and standardized TRUE/FALSE |
276 |
return(true); |
1
by brian
clean slate |
277 |
}
|
278 |
||
15
by brian
Fix for stat, NETWARE removal |
279 |
struct stat stat_info; |
280 |
if (stat(name,&stat_info)) |
|
51.1.53
by Jay Pipes
Removed/replaced DBUG symbols and standardized TRUE/FALSE |
281 |
return(true); |
1
by brian
clean slate |
282 |
|
283 |
// if we are not in slave thread, the file must be:
|
|
284 |
if (!thd->slave_thread && |
|
285 |
!((stat_info.st_mode & S_IROTH) == S_IROTH && // readable by others |
|
286 |
(stat_info.st_mode & S_IFLNK) != S_IFLNK && // and not a symlink |
|
287 |
((stat_info.st_mode & S_IFREG) == S_IFREG || |
|
288 |
(stat_info.st_mode & S_IFIFO) == S_IFIFO))) |
|
289 |
{
|
|
290 |
my_error(ER_TEXTFILE_NOT_READABLE, MYF(0), name); |
|
51.1.53
by Jay Pipes
Removed/replaced DBUG symbols and standardized TRUE/FALSE |
291 |
return(true); |
1
by brian
clean slate |
292 |
}
|
293 |
if ((stat_info.st_mode & S_IFIFO) == S_IFIFO) |
|
294 |
is_fifo = 1; |
|
295 |
}
|
|
296 |
if ((file=my_open(name,O_RDONLY,MYF(MY_WME))) < 0) |
|
51.1.53
by Jay Pipes
Removed/replaced DBUG symbols and standardized TRUE/FALSE |
297 |
return(true); |
1
by brian
clean slate |
298 |
}
|
299 |
||
300 |
COPY_INFO info; |
|
212.6.6
by Mats Kindahl
Removing redundant use of casts in drizzled/ for memcmp(), memcpy(), memset(), and memmove(). |
301 |
memset(&info, 0, sizeof(info)); |
1
by brian
clean slate |
302 |
info.ignore= ignore; |
303 |
info.handle_duplicates=handle_duplicates; |
|
304 |
info.escape_char=escaped->length() ? (*escaped)[0] : INT_MAX; |
|
305 |
||
306 |
READ_INFO read_info(file,tot_length, |
|
307 |
ex->cs ? ex->cs : thd->variables.collation_database, |
|
308 |
*field_term,*ex->line_start, *ex->line_term, *enclosed, |
|
309 |
info.escape_char, read_file_from_client, is_fifo); |
|
310 |
if (read_info.error) |
|
311 |
{
|
|
312 |
if (file >= 0) |
|
313 |
my_close(file,MYF(0)); // no files in net reading |
|
51.1.53
by Jay Pipes
Removed/replaced DBUG symbols and standardized TRUE/FALSE |
314 |
return(true); // Can't allocate buffers |
1
by brian
clean slate |
315 |
}
|
316 |
||
317 |
if (mysql_bin_log.is_open()) |
|
318 |
{
|
|
319 |
lf_info.thd = thd; |
|
320 |
lf_info.wrote_create_file = 0; |
|
321 |
lf_info.last_pos_in_file = HA_POS_ERROR; |
|
322 |
lf_info.log_delayed= transactional_table; |
|
323 |
read_info.set_io_cache_arg((void*) &lf_info); |
|
324 |
}
|
|
325 |
||
326 |
thd->count_cuted_fields= CHECK_FIELD_WARN; /* calc cuted fields */ |
|
327 |
thd->cuted_fields=0L; |
|
328 |
/* Skip lines if there is a line terminator */
|
|
266.1.23
by Monty Taylor
Removed load xml infile. Hope nobody liked it. Now the only thing we need xml.c |
329 |
if (ex->line_term->length()) |
1
by brian
clean slate |
330 |
{
|
331 |
/* ex->skip_lines needs to be preserved for logging */
|
|
332 |
while (skip_lines > 0) |
|
333 |
{
|
|
334 |
skip_lines--; |
|
335 |
if (read_info.next_line()) |
|
336 |
break; |
|
337 |
}
|
|
338 |
}
|
|
339 |
||
340 |
if (!(error=test(read_info.error))) |
|
341 |
{
|
|
342 |
||
343 |
table->next_number_field=table->found_next_number_field; |
|
344 |
if (ignore || |
|
345 |
handle_duplicates == DUP_REPLACE) |
|
346 |
table->file->extra(HA_EXTRA_IGNORE_DUP_KEY); |
|
347 |
if (handle_duplicates == DUP_REPLACE) |
|
348 |
table->file->extra(HA_EXTRA_WRITE_CAN_REPLACE); |
|
349 |
table->file->ha_start_bulk_insert((ha_rows) 0); |
|
350 |
table->copy_blobs=1; |
|
351 |
||
360
by Brian Aker
More MODE removal. |
352 |
thd->abort_on_warning= true; |
1
by brian
clean slate |
353 |
|
266.1.23
by Monty Taylor
Removed load xml infile. Hope nobody liked it. Now the only thing we need xml.c |
354 |
if (!field_term->length() && !enclosed->length()) |
1
by brian
clean slate |
355 |
error= read_fixed_length(thd, info, table_list, fields_vars, |
356 |
set_fields, set_values, read_info, |
|
357 |
skip_lines, ignore); |
|
358 |
else
|
|
359 |
error= read_sep_field(thd, info, table_list, fields_vars, |
|
360 |
set_fields, set_values, read_info, |
|
361 |
*enclosed, skip_lines, ignore); |
|
362 |
if (table->file->ha_end_bulk_insert() && !error) |
|
363 |
{
|
|
364 |
table->file->print_error(my_errno, MYF(0)); |
|
365 |
error= 1; |
|
366 |
}
|
|
367 |
table->file->extra(HA_EXTRA_NO_IGNORE_DUP_KEY); |
|
368 |
table->file->extra(HA_EXTRA_WRITE_CANNOT_REPLACE); |
|
369 |
table->next_number_field=0; |
|
370 |
}
|
|
371 |
if (file >= 0) |
|
372 |
my_close(file,MYF(0)); |
|
373 |
free_blobs(table); /* if pack_blob was used */ |
|
374 |
table->copy_blobs=0; |
|
375 |
thd->count_cuted_fields= CHECK_FIELD_IGNORE; |
|
376 |
/*
|
|
377 |
simulated killing in the middle of per-row loop
|
|
378 |
must be effective for binlogging
|
|
379 |
*/
|
|
380 |
killed_status= (error == 0)? THD::NOT_KILLED : thd->killed; |
|
381 |
if (error) |
|
382 |
{
|
|
383 |
if (read_file_from_client) |
|
384 |
while (!read_info.next_line()) |
|
385 |
;
|
|
386 |
||
387 |
if (mysql_bin_log.is_open()) |
|
388 |
{
|
|
389 |
{
|
|
390 |
/*
|
|
391 |
Make sure last block (the one which caused the error) gets
|
|
392 |
logged. This is needed because otherwise after write of (to
|
|
393 |
the binlog, not to read_info (which is a cache))
|
|
394 |
Delete_file_log_event the bad block will remain in read_info
|
|
395 |
(because pre_read is not called at the end of the last
|
|
396 |
block; remember pre_read is called whenever a new block is
|
|
397 |
read from disk). At the end of mysql_load(), the destructor
|
|
398 |
of read_info will call end_io_cache() which will flush
|
|
399 |
read_info, so we will finally have this in the binlog:
|
|
400 |
||
401 |
Append_block # The last successfull block
|
|
402 |
Delete_file
|
|
403 |
Append_block # The failing block
|
|
404 |
which is nonsense.
|
|
405 |
Or could also be (for a small file)
|
|
406 |
Create_file # The failing block
|
|
407 |
which is nonsense (Delete_file is not written in this case, because:
|
|
408 |
Create_file has not been written, so Delete_file is not written, then
|
|
409 |
when read_info is destroyed end_io_cache() is called which writes
|
|
410 |
Create_file.
|
|
411 |
*/
|
|
412 |
read_info.end_io_cache(); |
|
413 |
/* If the file was not empty, wrote_create_file is true */
|
|
414 |
if (lf_info.wrote_create_file) |
|
415 |
{
|
|
416 |
if (thd->transaction.stmt.modified_non_trans_table) |
|
417 |
write_execute_load_query_log_event(thd, handle_duplicates, |
|
418 |
ignore, transactional_table, |
|
419 |
killed_status); |
|
420 |
else
|
|
421 |
{
|
|
422 |
Delete_file_log_event d(thd, db, transactional_table); |
|
423 |
d.flags|= LOG_EVENT_UPDATE_TABLE_MAP_VERSION_F; |
|
424 |
mysql_bin_log.write(&d); |
|
425 |
}
|
|
426 |
}
|
|
427 |
}
|
|
428 |
}
|
|
429 |
error= -1; // Error on read |
|
430 |
goto err; |
|
431 |
}
|
|
307
by Brian Aker
Minor cleanups around ulong in kernel. |
432 |
sprintf(name, ER(ER_LOAD_INFO), (uint32_t) info.records, (uint32_t) info.deleted, |
433 |
(uint32_t) (info.records - info.copied), (uint32_t) thd->cuted_fields); |
|
1
by brian
clean slate |
434 |
|
435 |
if (thd->transaction.stmt.modified_non_trans_table) |
|
51.1.53
by Jay Pipes
Removed/replaced DBUG symbols and standardized TRUE/FALSE |
436 |
thd->transaction.all.modified_non_trans_table= true; |
1
by brian
clean slate |
437 |
|
438 |
if (mysql_bin_log.is_open()) |
|
439 |
{
|
|
440 |
/*
|
|
441 |
We need to do the job that is normally done inside
|
|
442 |
binlog_query() here, which is to ensure that the pending event
|
|
443 |
is written before tables are unlocked and before any other
|
|
444 |
events are written. We also need to update the table map
|
|
445 |
version for the binary log to mark that table maps are invalid
|
|
446 |
after this point.
|
|
447 |
*/
|
|
448 |
if (thd->current_stmt_binlog_row_based) |
|
449 |
thd->binlog_flush_pending_rows_event(true); |
|
450 |
else
|
|
451 |
{
|
|
452 |
/*
|
|
453 |
As already explained above, we need to call end_io_cache() or the last
|
|
454 |
block will be logged only after Execute_load_query_log_event (which is
|
|
455 |
wrong), when read_info is destroyed.
|
|
456 |
*/
|
|
457 |
read_info.end_io_cache(); |
|
458 |
if (lf_info.wrote_create_file) |
|
459 |
{
|
|
460 |
write_execute_load_query_log_event(thd, handle_duplicates, ignore, |
|
461 |
transactional_table,killed_status); |
|
462 |
}
|
|
463 |
}
|
|
464 |
}
|
|
465 |
||
466 |
/* ok to client sent only after binlog write and engine commit */
|
|
467 |
my_ok(thd, info.copied + info.deleted, 0L, name); |
|
468 |
err: |
|
51.1.53
by Jay Pipes
Removed/replaced DBUG symbols and standardized TRUE/FALSE |
469 |
assert(transactional_table || !(info.copied || info.deleted) || |
1
by brian
clean slate |
470 |
thd->transaction.stmt.modified_non_trans_table); |
471 |
table->file->ha_release_auto_increment(); |
|
51.1.53
by Jay Pipes
Removed/replaced DBUG symbols and standardized TRUE/FALSE |
472 |
table->auto_increment_field_not_null= false; |
1
by brian
clean slate |
473 |
thd->abort_on_warning= 0; |
51.1.1
by Jay Pipes
Merged PatG's removal of various DBUG stuff with still keeping DBUG_ASSERT calls since they seem to be breaking test runs |
474 |
return(error); |
1
by brian
clean slate |
475 |
}
|
476 |
||
477 |
||
478 |
/* Not a very useful function; just to avoid duplication of code */
|
|
479 |
static bool write_execute_load_query_log_event(THD *thd, |
|
480 |
bool duplicates, bool ignore, |
|
481 |
bool transactional_table, |
|
482 |
THD::killed_state killed_err_arg) |
|
483 |
{
|
|
484 |
Execute_load_query_log_event
|
|
485 |
e(thd, thd->query, thd->query_length, |
|
486 |
(char*)thd->lex->fname_start - (char*)thd->query, |
|
487 |
(char*)thd->lex->fname_end - (char*)thd->query, |
|
488 |
(duplicates == DUP_REPLACE) ? LOAD_DUP_REPLACE : |
|
489 |
(ignore ? LOAD_DUP_IGNORE : LOAD_DUP_ERROR), |
|
51.1.53
by Jay Pipes
Removed/replaced DBUG symbols and standardized TRUE/FALSE |
490 |
transactional_table, false, killed_err_arg); |
1
by brian
clean slate |
491 |
e.flags|= LOG_EVENT_UPDATE_TABLE_MAP_VERSION_F; |
492 |
return mysql_bin_log.write(&e); |
|
493 |
}
|
|
494 |
||
495 |
||
496 |
/****************************************************************************
|
|
497 |
** Read of rows of fixed size + optional garage + optonal newline
|
|
498 |
****************************************************************************/
|
|
499 |
||
500 |
static int |
|
327.2.4
by Brian Aker
Refactoring table.h |
501 |
read_fixed_length(THD *thd, COPY_INFO &info, TableList *table_list, |
1
by brian
clean slate |
502 |
List<Item> &fields_vars, List<Item> &set_fields, |
503 |
List<Item> &set_values, READ_INFO &read_info, |
|
307
by Brian Aker
Minor cleanups around ulong in kernel. |
504 |
uint32_t skip_lines, bool ignore_check_option_errors) |
1
by brian
clean slate |
505 |
{
|
506 |
List_iterator_fast<Item> it(fields_vars); |
|
507 |
Item_field *sql_field; |
|
327.1.5
by Brian Aker
Refactor around classes. TABLE_LIST has been factored out of table.h |
508 |
Table *table= table_list->table; |
151
by Brian Aker
Ulonglong to uint64_t |
509 |
uint64_t id; |
1
by brian
clean slate |
510 |
bool err; |
511 |
||
512 |
id= 0; |
|
513 |
||
514 |
while (!read_info.read_fixed_length()) |
|
515 |
{
|
|
516 |
if (thd->killed) |
|
517 |
{
|
|
518 |
thd->send_kill_message(); |
|
51.1.1
by Jay Pipes
Merged PatG's removal of various DBUG stuff with still keeping DBUG_ASSERT calls since they seem to be breaking test runs |
519 |
return(1); |
1
by brian
clean slate |
520 |
}
|
521 |
if (skip_lines) |
|
522 |
{
|
|
523 |
/*
|
|
524 |
We could implement this with a simple seek if:
|
|
525 |
- We are not using DATA INFILE LOCAL
|
|
526 |
- escape character is ""
|
|
527 |
- line starting prefix is ""
|
|
528 |
*/
|
|
529 |
skip_lines--; |
|
530 |
continue; |
|
531 |
}
|
|
532 |
it.rewind(); |
|
533 |
uchar *pos=read_info.row_start; |
|
534 |
#ifdef HAVE_purify
|
|
535 |
read_info.row_end[0]=0; |
|
536 |
#endif
|
|
537 |
||
538 |
restore_record(table, s->default_values); |
|
539 |
/*
|
|
540 |
There is no variables in fields_vars list in this format so
|
|
541 |
this conversion is safe.
|
|
542 |
*/
|
|
543 |
while ((sql_field= (Item_field*) it++)) |
|
544 |
{
|
|
545 |
Field *field= sql_field->field; |
|
546 |
if (field == table->next_number_field) |
|
51.1.53
by Jay Pipes
Removed/replaced DBUG symbols and standardized TRUE/FALSE |
547 |
table->auto_increment_field_not_null= true; |
1
by brian
clean slate |
548 |
/*
|
549 |
No fields specified in fields_vars list can be null in this format.
|
|
550 |
Mark field as not null, we should do this for each row because of
|
|
551 |
restore_record...
|
|
552 |
*/
|
|
553 |
field->set_notnull(); |
|
554 |
||
555 |
if (pos == read_info.row_end) |
|
556 |
{
|
|
557 |
thd->cuted_fields++; /* Not enough fields */ |
|
261.4.1
by Felipe
- Renamed MYSQL_ERROR to DRIZZLE_ERROR. |
558 |
push_warning_printf(thd, DRIZZLE_ERROR::WARN_LEVEL_WARN, |
1
by brian
clean slate |
559 |
ER_WARN_TOO_FEW_RECORDS, |
560 |
ER(ER_WARN_TOO_FEW_RECORDS), thd->row_count); |
|
212.2.2
by Patrick Galbraith
Renamed FIELD_TYPE to DRIZZLE_TYPE |
561 |
if (!field->maybe_null() && field->type() == DRIZZLE_TYPE_TIMESTAMP) |
1
by brian
clean slate |
562 |
((Field_timestamp*) field)->set_time(); |
563 |
}
|
|
564 |
else
|
|
565 |
{
|
|
566 |
uint length; |
|
567 |
uchar save_chr; |
|
568 |
if ((length=(uint) (read_info.row_end-pos)) > |
|
569 |
field->field_length) |
|
570 |
length=field->field_length; |
|
571 |
save_chr=pos[length]; pos[length]='\0'; // Safeguard aganst malloc |
|
572 |
field->store((char*) pos,length,read_info.read_charset); |
|
573 |
pos[length]=save_chr; |
|
574 |
if ((pos+=length) > read_info.row_end) |
|
575 |
pos= read_info.row_end; /* Fills rest with space */ |
|
576 |
}
|
|
577 |
}
|
|
578 |
if (pos != read_info.row_end) |
|
579 |
{
|
|
580 |
thd->cuted_fields++; /* To long row */ |
|
261.4.1
by Felipe
- Renamed MYSQL_ERROR to DRIZZLE_ERROR. |
581 |
push_warning_printf(thd, DRIZZLE_ERROR::WARN_LEVEL_WARN, |
1
by brian
clean slate |
582 |
ER_WARN_TOO_MANY_RECORDS, |
583 |
ER(ER_WARN_TOO_MANY_RECORDS), thd->row_count); |
|
584 |
}
|
|
585 |
||
586 |
if (thd->killed || |
|
587 |
fill_record(thd, set_fields, set_values, |
|
588 |
ignore_check_option_errors)) |
|
51.1.1
by Jay Pipes
Merged PatG's removal of various DBUG stuff with still keeping DBUG_ASSERT calls since they seem to be breaking test runs |
589 |
return(1); |
1
by brian
clean slate |
590 |
|
591 |
err= write_record(thd, table, &info); |
|
51.1.53
by Jay Pipes
Removed/replaced DBUG symbols and standardized TRUE/FALSE |
592 |
table->auto_increment_field_not_null= false; |
1
by brian
clean slate |
593 |
if (err) |
51.1.1
by Jay Pipes
Merged PatG's removal of various DBUG stuff with still keeping DBUG_ASSERT calls since they seem to be breaking test runs |
594 |
return(1); |
1
by brian
clean slate |
595 |
|
596 |
/*
|
|
597 |
We don't need to reset auto-increment field since we are restoring
|
|
598 |
its default value at the beginning of each loop iteration.
|
|
599 |
*/
|
|
600 |
if (read_info.next_line()) // Skip to next line |
|
601 |
break; |
|
602 |
if (read_info.line_cuted) |
|
603 |
{
|
|
604 |
thd->cuted_fields++; /* To long row */ |
|
261.4.1
by Felipe
- Renamed MYSQL_ERROR to DRIZZLE_ERROR. |
605 |
push_warning_printf(thd, DRIZZLE_ERROR::WARN_LEVEL_WARN, |
1
by brian
clean slate |
606 |
ER_WARN_TOO_MANY_RECORDS, |
607 |
ER(ER_WARN_TOO_MANY_RECORDS), thd->row_count); |
|
608 |
}
|
|
609 |
thd->row_count++; |
|
610 |
}
|
|
51.1.1
by Jay Pipes
Merged PatG's removal of various DBUG stuff with still keeping DBUG_ASSERT calls since they seem to be breaking test runs |
611 |
return(test(read_info.error)); |
1
by brian
clean slate |
612 |
}
|
613 |
||
614 |
||
615 |
||
616 |
static int |
|
327.2.4
by Brian Aker
Refactoring table.h |
617 |
read_sep_field(THD *thd, COPY_INFO &info, TableList *table_list, |
1
by brian
clean slate |
618 |
List<Item> &fields_vars, List<Item> &set_fields, |
619 |
List<Item> &set_values, READ_INFO &read_info, |
|
307
by Brian Aker
Minor cleanups around ulong in kernel. |
620 |
String &enclosed, uint32_t skip_lines, |
1
by brian
clean slate |
621 |
bool ignore_check_option_errors) |
622 |
{
|
|
623 |
List_iterator_fast<Item> it(fields_vars); |
|
624 |
Item *item; |
|
327.1.5
by Brian Aker
Refactor around classes. TABLE_LIST has been factored out of table.h |
625 |
Table *table= table_list->table; |
1
by brian
clean slate |
626 |
uint enclosed_length; |
151
by Brian Aker
Ulonglong to uint64_t |
627 |
uint64_t id; |
1
by brian
clean slate |
628 |
bool err; |
629 |
||
630 |
enclosed_length=enclosed.length(); |
|
631 |
id= 0; |
|
632 |
||
633 |
for (;;it.rewind()) |
|
634 |
{
|
|
635 |
if (thd->killed) |
|
636 |
{
|
|
637 |
thd->send_kill_message(); |
|
51.1.1
by Jay Pipes
Merged PatG's removal of various DBUG stuff with still keeping DBUG_ASSERT calls since they seem to be breaking test runs |
638 |
return(1); |
1
by brian
clean slate |
639 |
}
|
640 |
||
641 |
restore_record(table, s->default_values); |
|
642 |
||
643 |
while ((item= it++)) |
|
644 |
{
|
|
645 |
uint length; |
|
646 |
uchar *pos; |
|
647 |
Item *real_item; |
|
648 |
||
649 |
if (read_info.read_field()) |
|
650 |
break; |
|
651 |
||
652 |
/* If this line is to be skipped we don't want to fill field or var */
|
|
653 |
if (skip_lines) |
|
654 |
continue; |
|
655 |
||
656 |
pos=read_info.row_start; |
|
657 |
length=(uint) (read_info.row_end-pos); |
|
658 |
||
659 |
real_item= item->real_item(); |
|
660 |
||
661 |
if ((!read_info.enclosed && (enclosed_length && length == 4 && !memcmp(pos, STRING_WITH_LEN("NULL")))) || |
|
662 |
(length == 1 && read_info.found_null)) |
|
663 |
{
|
|
664 |
||
665 |
if (real_item->type() == Item::FIELD_ITEM) |
|
666 |
{
|
|
667 |
Field *field= ((Item_field *)real_item)->field; |
|
668 |
if (field->reset()) |
|
669 |
{
|
|
670 |
my_error(ER_WARN_NULL_TO_NOTNULL, MYF(0), field->field_name, |
|
671 |
thd->row_count); |
|
51.1.1
by Jay Pipes
Merged PatG's removal of various DBUG stuff with still keeping DBUG_ASSERT calls since they seem to be breaking test runs |
672 |
return(1); |
1
by brian
clean slate |
673 |
}
|
674 |
field->set_null(); |
|
675 |
if (!field->maybe_null()) |
|
676 |
{
|
|
212.2.2
by Patrick Galbraith
Renamed FIELD_TYPE to DRIZZLE_TYPE |
677 |
if (field->type() == DRIZZLE_TYPE_TIMESTAMP) |
1
by brian
clean slate |
678 |
((Field_timestamp*) field)->set_time(); |
679 |
else if (field != table->next_number_field) |
|
261.4.1
by Felipe
- Renamed MYSQL_ERROR to DRIZZLE_ERROR. |
680 |
field->set_warning(DRIZZLE_ERROR::WARN_LEVEL_WARN, |
1
by brian
clean slate |
681 |
ER_WARN_NULL_TO_NOTNULL, 1); |
682 |
}
|
|
683 |
}
|
|
684 |
else if (item->type() == Item::STRING_ITEM) |
|
685 |
{
|
|
686 |
((Item_user_var_as_out_param *)item)->set_null_value( |
|
687 |
read_info.read_charset); |
|
688 |
}
|
|
689 |
else
|
|
690 |
{
|
|
691 |
my_error(ER_LOAD_DATA_INVALID_COLUMN, MYF(0), item->full_name()); |
|
51.1.1
by Jay Pipes
Merged PatG's removal of various DBUG stuff with still keeping DBUG_ASSERT calls since they seem to be breaking test runs |
692 |
return(1); |
1
by brian
clean slate |
693 |
}
|
694 |
||
695 |
continue; |
|
696 |
}
|
|
697 |
||
698 |
if (real_item->type() == Item::FIELD_ITEM) |
|
699 |
{
|
|
700 |
Field *field= ((Item_field *)real_item)->field; |
|
701 |
field->set_notnull(); |
|
702 |
read_info.row_end[0]=0; // Safe to change end marker |
|
703 |
if (field == table->next_number_field) |
|
51.1.53
by Jay Pipes
Removed/replaced DBUG symbols and standardized TRUE/FALSE |
704 |
table->auto_increment_field_not_null= true; |
1
by brian
clean slate |
705 |
field->store((char*) pos, length, read_info.read_charset); |
706 |
}
|
|
707 |
else if (item->type() == Item::STRING_ITEM) |
|
708 |
{
|
|
709 |
((Item_user_var_as_out_param *)item)->set_value((char*) pos, length, |
|
710 |
read_info.read_charset); |
|
711 |
}
|
|
712 |
else
|
|
713 |
{
|
|
714 |
my_error(ER_LOAD_DATA_INVALID_COLUMN, MYF(0), item->full_name()); |
|
51.1.1
by Jay Pipes
Merged PatG's removal of various DBUG stuff with still keeping DBUG_ASSERT calls since they seem to be breaking test runs |
715 |
return(1); |
1
by brian
clean slate |
716 |
}
|
717 |
}
|
|
718 |
if (read_info.error) |
|
719 |
break; |
|
720 |
if (skip_lines) |
|
721 |
{
|
|
722 |
skip_lines--; |
|
723 |
continue; |
|
724 |
}
|
|
725 |
if (item) |
|
726 |
{
|
|
727 |
/* Have not read any field, thus input file is simply ended */
|
|
728 |
if (item == fields_vars.head()) |
|
729 |
break; |
|
730 |
for (; item ; item= it++) |
|
731 |
{
|
|
732 |
Item *real_item= item->real_item(); |
|
733 |
if (real_item->type() == Item::FIELD_ITEM) |
|
734 |
{
|
|
735 |
Field *field= ((Item_field *)real_item)->field; |
|
736 |
if (field->reset()) |
|
737 |
{
|
|
738 |
my_error(ER_WARN_NULL_TO_NOTNULL, MYF(0),field->field_name, |
|
739 |
thd->row_count); |
|
51.1.1
by Jay Pipes
Merged PatG's removal of various DBUG stuff with still keeping DBUG_ASSERT calls since they seem to be breaking test runs |
740 |
return(1); |
1
by brian
clean slate |
741 |
}
|
212.2.2
by Patrick Galbraith
Renamed FIELD_TYPE to DRIZZLE_TYPE |
742 |
if (!field->maybe_null() && field->type() == DRIZZLE_TYPE_TIMESTAMP) |
1
by brian
clean slate |
743 |
((Field_timestamp*) field)->set_time(); |
744 |
/*
|
|
745 |
QQ: We probably should not throw warning for each field.
|
|
746 |
But how about intention to always have the same number
|
|
747 |
of warnings in THD::cuted_fields (and get rid of cuted_fields
|
|
748 |
in the end ?)
|
|
749 |
*/
|
|
750 |
thd->cuted_fields++; |
|
261.4.1
by Felipe
- Renamed MYSQL_ERROR to DRIZZLE_ERROR. |
751 |
push_warning_printf(thd, DRIZZLE_ERROR::WARN_LEVEL_WARN, |
1
by brian
clean slate |
752 |
ER_WARN_TOO_FEW_RECORDS, |
753 |
ER(ER_WARN_TOO_FEW_RECORDS), thd->row_count); |
|
754 |
}
|
|
755 |
else if (item->type() == Item::STRING_ITEM) |
|
756 |
{
|
|
757 |
((Item_user_var_as_out_param *)item)->set_null_value( |
|
758 |
read_info.read_charset); |
|
759 |
}
|
|
760 |
else
|
|
761 |
{
|
|
762 |
my_error(ER_LOAD_DATA_INVALID_COLUMN, MYF(0), item->full_name()); |
|
51.1.1
by Jay Pipes
Merged PatG's removal of various DBUG stuff with still keeping DBUG_ASSERT calls since they seem to be breaking test runs |
763 |
return(1); |
1
by brian
clean slate |
764 |
}
|
765 |
}
|
|
766 |
}
|
|
767 |
||
768 |
if (thd->killed || |
|
769 |
fill_record(thd, set_fields, set_values, |
|
770 |
ignore_check_option_errors)) |
|
51.1.1
by Jay Pipes
Merged PatG's removal of various DBUG stuff with still keeping DBUG_ASSERT calls since they seem to be breaking test runs |
771 |
return(1); |
1
by brian
clean slate |
772 |
|
773 |
err= write_record(thd, table, &info); |
|
51.1.53
by Jay Pipes
Removed/replaced DBUG symbols and standardized TRUE/FALSE |
774 |
table->auto_increment_field_not_null= false; |
1
by brian
clean slate |
775 |
if (err) |
51.1.1
by Jay Pipes
Merged PatG's removal of various DBUG stuff with still keeping DBUG_ASSERT calls since they seem to be breaking test runs |
776 |
return(1); |
1
by brian
clean slate |
777 |
/*
|
778 |
We don't need to reset auto-increment field since we are restoring
|
|
779 |
its default value at the beginning of each loop iteration.
|
|
780 |
*/
|
|
781 |
if (read_info.next_line()) // Skip to next line |
|
782 |
break; |
|
783 |
if (read_info.line_cuted) |
|
784 |
{
|
|
785 |
thd->cuted_fields++; /* To long row */ |
|
261.4.1
by Felipe
- Renamed MYSQL_ERROR to DRIZZLE_ERROR. |
786 |
push_warning_printf(thd, DRIZZLE_ERROR::WARN_LEVEL_WARN, |
1
by brian
clean slate |
787 |
ER_WARN_TOO_MANY_RECORDS, ER(ER_WARN_TOO_MANY_RECORDS), |
788 |
thd->row_count); |
|
789 |
if (thd->killed) |
|
51.1.1
by Jay Pipes
Merged PatG's removal of various DBUG stuff with still keeping DBUG_ASSERT calls since they seem to be breaking test runs |
790 |
return(1); |
1
by brian
clean slate |
791 |
}
|
792 |
thd->row_count++; |
|
793 |
}
|
|
51.1.1
by Jay Pipes
Merged PatG's removal of various DBUG stuff with still keeping DBUG_ASSERT calls since they seem to be breaking test runs |
794 |
return(test(read_info.error)); |
1
by brian
clean slate |
795 |
}
|
796 |
||
797 |
||
798 |
/* Unescape all escape characters, mark \N as null */
|
|
799 |
||
800 |
char
|
|
801 |
READ_INFO::unescape(char chr) |
|
802 |
{
|
|
803 |
/* keep this switch synchornous with the ESCAPE_CHARS macro */
|
|
804 |
switch(chr) { |
|
805 |
case 'n': return '\n'; |
|
806 |
case 't': return '\t'; |
|
807 |
case 'r': return '\r'; |
|
808 |
case 'b': return '\b'; |
|
809 |
case '0': return 0; // Ascii null |
|
810 |
case 'Z': return '\032'; // Win32 end of file |
|
811 |
case 'N': found_null=1; |
|
812 |
||
813 |
/* fall through */
|
|
814 |
default: return chr; |
|
815 |
}
|
|
816 |
}
|
|
817 |
||
818 |
||
819 |
/*
|
|
820 |
Read a line using buffering
|
|
821 |
If last line is empty (in line mode) then it isn't outputed
|
|
822 |
*/
|
|
823 |
||
824 |
||
264.2.6
by Andrey Hristov
Constify the usage of CHARSET_INFO almost to the last place in the code. |
825 |
READ_INFO::READ_INFO(File file_par, uint tot_length, const CHARSET_INFO * const cs, |
1
by brian
clean slate |
826 |
String &field_term, String &line_start, String &line_term, |
827 |
String &enclosed_par, int escape, bool get_it_from_net, |
|
828 |
bool is_fifo) |
|
829 |
:file(file_par),escape_char(escape) |
|
830 |
{
|
|
831 |
read_charset= cs; |
|
832 |
field_term_ptr=(char*) field_term.ptr(); |
|
833 |
field_term_length= field_term.length(); |
|
834 |
line_term_ptr=(char*) line_term.ptr(); |
|
835 |
line_term_length= line_term.length(); |
|
836 |
if (line_start.length() == 0) |
|
837 |
{
|
|
838 |
line_start_ptr=0; |
|
839 |
start_of_line= 0; |
|
840 |
}
|
|
841 |
else
|
|
842 |
{
|
|
843 |
line_start_ptr=(char*) line_start.ptr(); |
|
844 |
line_start_end=line_start_ptr+line_start.length(); |
|
845 |
start_of_line= 1; |
|
846 |
}
|
|
847 |
/* If field_terminator == line_terminator, don't use line_terminator */
|
|
848 |
if (field_term_length == line_term_length && |
|
849 |
!memcmp(field_term_ptr,line_term_ptr,field_term_length)) |
|
850 |
{
|
|
851 |
line_term_length=0; |
|
852 |
line_term_ptr=(char*) ""; |
|
853 |
}
|
|
854 |
enclosed_char= (enclosed_length=enclosed_par.length()) ? |
|
855 |
(uchar) enclosed_par[0] : INT_MAX; |
|
856 |
field_term_char= field_term_length ? (uchar) field_term_ptr[0] : INT_MAX; |
|
857 |
line_term_char= line_term_length ? (uchar) line_term_ptr[0] : INT_MAX; |
|
858 |
error=eof=found_end_of_line=found_null=line_cuted=0; |
|
859 |
buff_length=tot_length; |
|
860 |
||
861 |
||
862 |
/* Set of a stack for unget if long terminators */
|
|
398.1.4
by Monty Taylor
Renamed max/min. |
863 |
uint length=cmax(field_term_length,line_term_length)+1; |
1
by brian
clean slate |
864 |
set_if_bigger(length,line_start.length()); |
865 |
stack=stack_pos=(int*) sql_alloc(sizeof(int)*length); |
|
866 |
||
867 |
if (!(buffer=(uchar*) my_malloc(buff_length+1,MYF(0)))) |
|
868 |
error=1; /* purecov: inspected */ |
|
869 |
else
|
|
870 |
{
|
|
871 |
end_of_buff=buffer+buff_length; |
|
872 |
if (init_io_cache(&cache,(get_it_from_net) ? -1 : file, 0, |
|
873 |
(get_it_from_net) ? READ_NET : |
|
874 |
(is_fifo ? READ_FIFO : READ_CACHE),0L,1, |
|
875 |
MYF(MY_WME))) |
|
876 |
{
|
|
877 |
my_free((uchar*) buffer,MYF(0)); /* purecov: inspected */ |
|
878 |
error=1; |
|
879 |
}
|
|
880 |
else
|
|
881 |
{
|
|
882 |
/*
|
|
883 |
init_io_cache() will not initialize read_function member
|
|
884 |
if the cache is READ_NET. So we work around the problem with a
|
|
885 |
manual assignment
|
|
886 |
*/
|
|
887 |
need_end_io_cache = 1; |
|
888 |
||
889 |
if (get_it_from_net) |
|
890 |
cache.read_function = _my_b_net_read; |
|
891 |
||
892 |
if (mysql_bin_log.is_open()) |
|
893 |
cache.pre_read = cache.pre_close = |
|
894 |
(IO_CACHE_CALLBACK) log_loaded_block; |
|
895 |
}
|
|
896 |
}
|
|
897 |
}
|
|
898 |
||
899 |
||
900 |
READ_INFO::~READ_INFO() |
|
901 |
{
|
|
902 |
if (!error) |
|
903 |
{
|
|
904 |
if (need_end_io_cache) |
|
905 |
::end_io_cache(&cache); |
|
906 |
my_free((uchar*) buffer,MYF(0)); |
|
907 |
error=1; |
|
908 |
}
|
|
909 |
}
|
|
910 |
||
911 |
||
912 |
#define GET (stack_pos != stack ? *--stack_pos : my_b_get(&cache))
|
|
913 |
#define PUSH(A) *(stack_pos++)=(A)
|
|
914 |
||
915 |
||
916 |
inline int READ_INFO::terminator(char *ptr,uint length) |
|
917 |
{
|
|
918 |
int chr=0; // Keep gcc happy |
|
919 |
uint i; |
|
920 |
for (i=1 ; i < length ; i++) |
|
921 |
{
|
|
922 |
if ((chr=GET) != *++ptr) |
|
923 |
{
|
|
924 |
break; |
|
925 |
}
|
|
926 |
}
|
|
927 |
if (i == length) |
|
928 |
return 1; |
|
929 |
PUSH(chr); |
|
930 |
while (i-- > 1) |
|
931 |
PUSH((uchar) *--ptr); |
|
932 |
return 0; |
|
933 |
}
|
|
934 |
||
935 |
||
936 |
int READ_INFO::read_field() |
|
937 |
{
|
|
938 |
int chr,found_enclosed_char; |
|
939 |
uchar *to,*new_buffer; |
|
940 |
||
941 |
found_null=0; |
|
942 |
if (found_end_of_line) |
|
943 |
return 1; // One have to call next_line |
|
944 |
||
945 |
/* Skip until we find 'line_start' */
|
|
946 |
||
947 |
if (start_of_line) |
|
948 |
{ // Skip until line_start |
|
949 |
start_of_line=0; |
|
950 |
if (find_start_of_fields()) |
|
951 |
return 1; |
|
952 |
}
|
|
953 |
if ((chr=GET) == my_b_EOF) |
|
954 |
{
|
|
955 |
found_end_of_line=eof=1; |
|
956 |
return 1; |
|
957 |
}
|
|
958 |
to=buffer; |
|
959 |
if (chr == enclosed_char) |
|
960 |
{
|
|
961 |
found_enclosed_char=enclosed_char; |
|
962 |
*to++=(uchar) chr; // If error |
|
963 |
}
|
|
964 |
else
|
|
965 |
{
|
|
966 |
found_enclosed_char= INT_MAX; |
|
967 |
PUSH(chr); |
|
968 |
}
|
|
969 |
||
970 |
for (;;) |
|
971 |
{
|
|
972 |
while ( to < end_of_buff) |
|
973 |
{
|
|
974 |
chr = GET; |
|
975 |
#ifdef USE_MB
|
|
976 |
if ((my_mbcharlen(read_charset, chr) > 1) && |
|
977 |
to+my_mbcharlen(read_charset, chr) <= end_of_buff) |
|
978 |
{
|
|
979 |
uchar* p = (uchar*)to; |
|
980 |
*to++ = chr; |
|
981 |
int ml = my_mbcharlen(read_charset, chr); |
|
982 |
int i; |
|
983 |
for (i=1; i<ml; i++) { |
|
984 |
chr = GET; |
|
985 |
if (chr == my_b_EOF) |
|
986 |
goto found_eof; |
|
987 |
*to++ = chr; |
|
988 |
}
|
|
989 |
if (my_ismbchar(read_charset, |
|
990 |
(const char *)p, |
|
991 |
(const char *)to)) |
|
992 |
continue; |
|
993 |
for (i=0; i<ml; i++) |
|
994 |
PUSH((uchar) *--to); |
|
995 |
chr = GET; |
|
996 |
}
|
|
997 |
#endif
|
|
998 |
if (chr == my_b_EOF) |
|
999 |
goto found_eof; |
|
1000 |
if (chr == escape_char) |
|
1001 |
{
|
|
1002 |
if ((chr=GET) == my_b_EOF) |
|
1003 |
{
|
|
1004 |
*to++= (uchar) escape_char; |
|
1005 |
goto found_eof; |
|
1006 |
}
|
|
1007 |
/*
|
|
1008 |
When escape_char == enclosed_char, we treat it like we do for
|
|
1009 |
handling quotes in SQL parsing -- you can double-up the
|
|
1010 |
escape_char to include it literally, but it doesn't do escapes
|
|
1011 |
like \n. This allows: LOAD DATA ... ENCLOSED BY '"' ESCAPED BY '"'
|
|
1012 |
with data like: "fie""ld1", "field2"
|
|
1013 |
*/
|
|
1014 |
if (escape_char != enclosed_char || chr == escape_char) |
|
1015 |
{
|
|
1016 |
*to++ = (uchar) unescape((char) chr); |
|
1017 |
continue; |
|
1018 |
}
|
|
1019 |
PUSH(chr); |
|
1020 |
chr= escape_char; |
|
1021 |
}
|
|
1022 |
#ifdef ALLOW_LINESEPARATOR_IN_STRINGS
|
|
1023 |
if (chr == line_term_char) |
|
1024 |
#else
|
|
1025 |
if (chr == line_term_char && found_enclosed_char == INT_MAX) |
|
1026 |
#endif
|
|
1027 |
{
|
|
1028 |
if (terminator(line_term_ptr,line_term_length)) |
|
1029 |
{ // Maybe unexpected linefeed |
|
1030 |
enclosed=0; |
|
1031 |
found_end_of_line=1; |
|
1032 |
row_start=buffer; |
|
1033 |
row_end= to; |
|
1034 |
return 0; |
|
1035 |
}
|
|
1036 |
}
|
|
1037 |
if (chr == found_enclosed_char) |
|
1038 |
{
|
|
1039 |
if ((chr=GET) == found_enclosed_char) |
|
1040 |
{ // Remove dupplicated |
|
1041 |
*to++ = (uchar) chr; |
|
1042 |
continue; |
|
1043 |
}
|
|
1044 |
// End of enclosed field if followed by field_term or line_term
|
|
1045 |
if (chr == my_b_EOF || |
|
1046 |
(chr == line_term_char && terminator(line_term_ptr, line_term_length))) |
|
1047 |
{ // Maybe unexpected linefeed |
|
1048 |
enclosed=1; |
|
1049 |
found_end_of_line=1; |
|
1050 |
row_start=buffer+1; |
|
1051 |
row_end= to; |
|
1052 |
return 0; |
|
1053 |
}
|
|
1054 |
if (chr == field_term_char && |
|
1055 |
terminator(field_term_ptr,field_term_length)) |
|
1056 |
{
|
|
1057 |
enclosed=1; |
|
1058 |
row_start=buffer+1; |
|
1059 |
row_end= to; |
|
1060 |
return 0; |
|
1061 |
}
|
|
1062 |
/*
|
|
1063 |
The string didn't terminate yet.
|
|
1064 |
Store back next character for the loop
|
|
1065 |
*/
|
|
1066 |
PUSH(chr); |
|
1067 |
/* copy the found term character to 'to' */
|
|
1068 |
chr= found_enclosed_char; |
|
1069 |
}
|
|
1070 |
else if (chr == field_term_char && found_enclosed_char == INT_MAX) |
|
1071 |
{
|
|
1072 |
if (terminator(field_term_ptr,field_term_length)) |
|
1073 |
{
|
|
1074 |
enclosed=0; |
|
1075 |
row_start=buffer; |
|
1076 |
row_end= to; |
|
1077 |
return 0; |
|
1078 |
}
|
|
1079 |
}
|
|
1080 |
*to++ = (uchar) chr; |
|
1081 |
}
|
|
1082 |
/*
|
|
1083 |
** We come here if buffer is too small. Enlarge it and continue
|
|
1084 |
*/
|
|
1085 |
if (!(new_buffer=(uchar*) my_realloc((char*) buffer,buff_length+1+IO_SIZE, |
|
1086 |
MYF(MY_WME)))) |
|
1087 |
return (error=1); |
|
1088 |
to=new_buffer + (to-buffer); |
|
1089 |
buffer=new_buffer; |
|
1090 |
buff_length+=IO_SIZE; |
|
1091 |
end_of_buff=buffer+buff_length; |
|
1092 |
}
|
|
1093 |
||
1094 |
found_eof: |
|
1095 |
enclosed=0; |
|
1096 |
found_end_of_line=eof=1; |
|
1097 |
row_start=buffer; |
|
1098 |
row_end=to; |
|
1099 |
return 0; |
|
1100 |
}
|
|
1101 |
||
1102 |
/*
|
|
1103 |
Read a row with fixed length.
|
|
1104 |
||
1105 |
NOTES
|
|
1106 |
The row may not be fixed size on disk if there are escape
|
|
1107 |
characters in the file.
|
|
1108 |
||
1109 |
IMPLEMENTATION NOTE
|
|
1110 |
One can't use fixed length with multi-byte charset **
|
|
1111 |
||
1112 |
RETURN
|
|
1113 |
0 ok
|
|
1114 |
1 error
|
|
1115 |
*/
|
|
1116 |
||
1117 |
int READ_INFO::read_fixed_length() |
|
1118 |
{
|
|
1119 |
int chr; |
|
1120 |
uchar *to; |
|
1121 |
if (found_end_of_line) |
|
1122 |
return 1; // One have to call next_line |
|
1123 |
||
1124 |
if (start_of_line) |
|
1125 |
{ // Skip until line_start |
|
1126 |
start_of_line=0; |
|
1127 |
if (find_start_of_fields()) |
|
1128 |
return 1; |
|
1129 |
}
|
|
1130 |
||
1131 |
to=row_start=buffer; |
|
1132 |
while (to < end_of_buff) |
|
1133 |
{
|
|
1134 |
if ((chr=GET) == my_b_EOF) |
|
1135 |
goto found_eof; |
|
1136 |
if (chr == escape_char) |
|
1137 |
{
|
|
1138 |
if ((chr=GET) == my_b_EOF) |
|
1139 |
{
|
|
1140 |
*to++= (uchar) escape_char; |
|
1141 |
goto found_eof; |
|
1142 |
}
|
|
1143 |
*to++ =(uchar) unescape((char) chr); |
|
1144 |
continue; |
|
1145 |
}
|
|
1146 |
if (chr == line_term_char) |
|
1147 |
{
|
|
1148 |
if (terminator(line_term_ptr,line_term_length)) |
|
1149 |
{ // Maybe unexpected linefeed |
|
1150 |
found_end_of_line=1; |
|
1151 |
row_end= to; |
|
1152 |
return 0; |
|
1153 |
}
|
|
1154 |
}
|
|
1155 |
*to++ = (uchar) chr; |
|
1156 |
}
|
|
1157 |
row_end=to; // Found full line |
|
1158 |
return 0; |
|
1159 |
||
1160 |
found_eof: |
|
1161 |
found_end_of_line=eof=1; |
|
1162 |
row_start=buffer; |
|
1163 |
row_end=to; |
|
1164 |
return to == buffer ? 1 : 0; |
|
1165 |
}
|
|
1166 |
||
1167 |
||
1168 |
int READ_INFO::next_line() |
|
1169 |
{
|
|
1170 |
line_cuted=0; |
|
1171 |
start_of_line= line_start_ptr != 0; |
|
1172 |
if (found_end_of_line || eof) |
|
1173 |
{
|
|
1174 |
found_end_of_line=0; |
|
1175 |
return eof; |
|
1176 |
}
|
|
1177 |
found_end_of_line=0; |
|
1178 |
if (!line_term_length) |
|
1179 |
return 0; // No lines |
|
1180 |
for (;;) |
|
1181 |
{
|
|
1182 |
int chr = GET; |
|
1183 |
#ifdef USE_MB
|
|
1184 |
if (my_mbcharlen(read_charset, chr) > 1) |
|
1185 |
{
|
|
1186 |
for (uint i=1; |
|
1187 |
chr != my_b_EOF && i<my_mbcharlen(read_charset, chr); |
|
1188 |
i++) |
|
1189 |
chr = GET; |
|
1190 |
if (chr == escape_char) |
|
1191 |
continue; |
|
1192 |
}
|
|
1193 |
#endif
|
|
1194 |
if (chr == my_b_EOF) |
|
1195 |
{
|
|
1196 |
eof=1; |
|
1197 |
return 1; |
|
1198 |
}
|
|
1199 |
if (chr == escape_char) |
|
1200 |
{
|
|
1201 |
line_cuted=1; |
|
1202 |
if (GET == my_b_EOF) |
|
1203 |
return 1; |
|
1204 |
continue; |
|
1205 |
}
|
|
1206 |
if (chr == line_term_char && terminator(line_term_ptr,line_term_length)) |
|
1207 |
return 0; |
|
1208 |
line_cuted=1; |
|
1209 |
}
|
|
1210 |
}
|
|
1211 |
||
1212 |
||
1213 |
bool READ_INFO::find_start_of_fields() |
|
1214 |
{
|
|
1215 |
int chr; |
|
1216 |
try_again: |
|
1217 |
do
|
|
1218 |
{
|
|
1219 |
if ((chr=GET) == my_b_EOF) |
|
1220 |
{
|
|
1221 |
found_end_of_line=eof=1; |
|
1222 |
return 1; |
|
1223 |
}
|
|
1224 |
} while ((char) chr != line_start_ptr[0]); |
|
1225 |
for (char *ptr=line_start_ptr+1 ; ptr != line_start_end ; ptr++) |
|
1226 |
{
|
|
1227 |
chr=GET; // Eof will be checked later |
|
1228 |
if ((char) chr != *ptr) |
|
1229 |
{ // Can't be line_start |
|
1230 |
PUSH(chr); |
|
1231 |
while (--ptr != line_start_ptr) |
|
1232 |
{ // Restart with next char |
|
1233 |
PUSH((uchar) *ptr); |
|
1234 |
}
|
|
1235 |
goto try_again; |
|
1236 |
}
|
|
1237 |
}
|
|
1238 |
return 0; |
|
1239 |
}
|
|
1240 |
||
1241 |