4 use Carp ':DEFAULT', 'confess';
6 use Fcntl 'O_CREAT', 'O_RDWR', 'LOCK_EX', 'LOCK_SH', 'O_WRONLY', 'O_RDONLY';
7 sub O_ACCMODE () { O_RDONLY | O_RDWR | O_WRONLY }
11 my $DEFAULT_MEMORY_SIZE = 1<<21; # 2 megabytes
12 my $DEFAULT_AUTODEFER_THRESHHOLD = 3; # 3 records
13 my $DEFAULT_AUTODEFER_FILELEN_THRESHHOLD = 65536; # 16 disk blocksful
15 my %good_opt = map {$_ => 1, "-$_" => 1}
16 qw(memory dw_size mode recsep discipline
17 autodefer autochomp autodefer_threshhold concurrent);
21 croak "usage: tie \@array, $_[0], filename, [option => value]...";
23 my ($pack, $file, %opts) = @_;
25 # transform '-foo' keys into 'foo' keys
26 for my $key (keys %opts) {
27 unless ($good_opt{$key}) {
28 croak("$pack: Unrecognized option '$key'\n");
31 if ($key =~ s/^-+//) {
32 $opts{$key} = delete $opts{$okey};
36 if ($opts{concurrent}) {
37 croak("$pack: concurrent access not supported yet\n");
40 unless (defined $opts{memory}) {
41 # default is the larger of the default cache size and the
42 # deferred-write buffer size (if specified)
43 $opts{memory} = $DEFAULT_MEMORY_SIZE;
44 $opts{memory} = $opts{dw_size}
45 if defined $opts{dw_size} && $opts{dw_size} > $DEFAULT_MEMORY_SIZE;
48 $opts{dw_size} = $opts{memory} unless defined $opts{dw_size};
49 if ($opts{dw_size} > $opts{memory}) {
50 croak("$pack: dw_size may not be larger than total memory allocation\n");
52 # are we in deferred-write mode?
53 $opts{defer} = 0 unless defined $opts{defer};
54 $opts{deferred} = {}; # no records are presently deferred
55 $opts{deferred_s} = 0; # count of total bytes in ->{deferred}
56 $opts{deferred_max} = -1; # empty
58 # What's a good way to arrange that this class can be overridden?
59 $opts{cache} = Tie::File::Cache->new($opts{memory});
61 # autodeferment is enabled by default
62 $opts{autodefer} = 1 unless defined $opts{autodefer};
63 $opts{autodeferring} = 0; # but is not initially active
64 $opts{ad_history} = [];
65 $opts{autodefer_threshhold} = $DEFAULT_AUTODEFER_THRESHHOLD
66 unless defined $opts{autodefer_threshhold};
67 $opts{autodefer_filelen_threshhold} = $DEFAULT_AUTODEFER_FILELEN_THRESHHOLD
68 unless defined $opts{autodefer_filelen_threshhold};
71 $opts{filename} = $file;
72 unless (defined $opts{recsep}) {
73 $opts{recsep} = _default_recsep();
75 $opts{recseplen} = length($opts{recsep});
76 if ($opts{recseplen} == 0) {
77 croak "Empty record separator not supported by $pack";
80 $opts{autochomp} = 1 unless defined $opts{autochomp};
82 $opts{mode} = O_CREAT|O_RDWR unless defined $opts{mode};
83 $opts{rdonly} = (($opts{mode} & O_ACCMODE) == O_RDONLY);
84 $opts{sawlastrec} = undef;
88 if (UNIVERSAL::isa($file, 'GLOB')) {
89 # We use 1 here on the theory that some systems
90 # may not indicate failure if we use 0.
91 # MSWin32 does not indicate failure with 0, but I don't know if
92 # it will indicate failure with 1 or not.
93 unless (seek $file, 1, SEEK_SET) {
94 croak "$pack: your filehandle does not appear to be seekable";
96 seek $file, 0, SEEK_SET; # put it back
97 $fh = $file; # setting binmode is the user's problem
99 croak "usage: tie \@array, $pack, filename, [option => value]...";
101 # $fh = \do { local *FH }; # XXX this is buggy
103 # perl 5.005 and earlier don't autovivify filehandles
105 $fh = Symbol::gensym();
107 sysopen $fh, $file, $opts{mode}, 0666 or return;
111 { my $ofh = select $fh; $| = 1; select $ofh } # autoflush on write
112 if (defined $opts{discipline} && $] >= 5.006) {
113 # This avoids a compile-time warning under 5.005
114 eval 'binmode($fh, $opts{discipline})';
115 croak $@ if $@ =~ /unknown discipline/i;
120 bless \%opts => $pack;
127 # check the defer buffer
128 $rec = $self->{deferred}{$n} if exists $self->{deferred}{$n};
129 $rec = $self->_fetch($n) unless defined $rec;
132 substr($rec, - $self->{recseplen}) = ""
133 if defined $rec && $self->{autochomp};
137 # Chomp many records in-place; return nothing useful
140 return unless $self->{autochomp};
141 if ($self->{autochomp}) {
144 substr($_, - $self->{recseplen}) = "";
149 # Chomp one record in-place; return modified record
151 my ($self, $rec) = @_;
152 return $rec unless $self->{autochomp};
153 return unless defined $rec;
154 substr($rec, - $self->{recseplen}) = "";
161 # check the record cache
162 { my $cached = $self->{cache}->lookup($n);
163 return $cached if defined $cached;
166 if ($#{$self->{offsets}} < $n) {
167 return if $self->{eof}; # request for record beyond end of file
168 my $o = $self->_fill_offsets_to($n);
169 # If it's still undefined, there is no such record, so return 'undef'
170 return unless defined $o;
173 my $fh = $self->{FH};
174 $self->_seek($n); # we can do this now that offsets is populated
175 my $rec = $self->_read_record;
177 # If we happen to have just read the first record, check to see if
178 # the length of the record matches what 'tell' says. If not, Tie::File
179 # won't work, and should drop dead.
181 # if ($n == 0 && defined($rec) && tell($self->{fh}) != length($rec)) {
182 # if (defined $self->{discipline}) {
183 # croak "I/O discipline $self->{discipline} not supported";
185 # croak "File encoding not supported";
189 $self->{cache}->insert($n, $rec) if defined $rec && not $self->{flushing};
194 my ($self, $n, $rec) = @_;
195 die "STORE called from _check_integrity!" if $DIAGNOSTIC;
197 $self->_fixrecs($rec);
199 if ($self->{autodefer}) {
200 $self->_annotate_ad_history($n);
203 return $self->_store_deferred($n, $rec) if $self->_is_deferring;
206 # We need this to decide whether the new record will fit
207 # It incidentally populates the offsets table
208 # Note we have to do this before we alter the cache
209 # 20020324 Wait, but this DOES alter the cache. TODO BUG?
210 my $oldrec = $self->_fetch($n);
212 if (not defined $oldrec) {
213 # We're storing a record beyond the end of the file
214 $self->_extend_file_to($n+1);
215 $oldrec = $self->{recsep};
217 # return if $oldrec eq $rec; # don't bother
218 my $len_diff = length($rec) - length($oldrec);
220 # length($oldrec) here is not consistent with text mode TODO XXX BUG
221 $self->_mtwrite($rec, $self->{offsets}[$n], length($oldrec));
222 $self->_oadjust([$n, 1, $rec]);
223 $self->{cache}->update($n, $rec);
226 sub _store_deferred {
227 my ($self, $n, $rec) = @_;
228 $self->{cache}->remove($n);
229 my $old_deferred = $self->{deferred}{$n};
231 if (defined $self->{deferred_max} && $n > $self->{deferred_max}) {
232 $self->{deferred_max} = $n;
234 $self->{deferred}{$n} = $rec;
236 my $len_diff = length($rec);
237 $len_diff -= length($old_deferred) if defined $old_deferred;
238 $self->{deferred_s} += $len_diff;
239 $self->{cache}->adj_limit(-$len_diff);
240 if ($self->{deferred_s} > $self->{dw_size}) {
242 } elsif ($self->_cache_too_full) {
247 # Remove a single record from the deferred-write buffer without writing it
248 # The record need not be present
249 sub _delete_deferred {
251 my $rec = delete $self->{deferred}{$n};
252 return unless defined $rec;
254 if (defined $self->{deferred_max}
255 && $n == $self->{deferred_max}) {
256 undef $self->{deferred_max};
259 $self->{deferred_s} -= length $rec;
260 $self->{cache}->adj_limit(length $rec);
265 my $n = $self->{eof} ? $#{$self->{offsets}} : $self->_fill_offsets;
267 my $top_deferred = $self->_defer_max;
268 $n = $top_deferred+1 if defined $top_deferred && $n < $top_deferred+1;
273 my ($self, $len) = @_;
275 if ($self->{autodefer}) {
276 $self->_annotate_ad_history('STORESIZE');
279 my $olen = $self->FETCHSIZE;
280 return if $len == $olen; # Woo-hoo!
284 if ($self->_is_deferring) {
285 for ($olen .. $len-1) {
286 $self->_store_deferred($_, $self->{recsep});
289 $self->_extend_file_to($len);
295 if ($self->_is_deferring) {
296 # TODO maybe replace this with map-plus-assignment?
297 for (grep $_ >= $len, keys %{$self->{deferred}}) {
298 $self->_delete_deferred($_);
300 $self->{deferred_max} = $len-1;
305 $#{$self->{offsets}} = $len;
306 # $self->{offsets}[0] = 0; # in case we just chopped this
308 $self->{cache}->remove(grep $_ >= $len, $self->{cache}->ckeys);
312 ### It should not be necessary to do FETCHSIZE
313 ### Just seek to the end of the file.
316 $self->SPLICE($self->FETCHSIZE, scalar(@_), @_);
319 # $self->FETCHSIZE; # because av.c takes care of this for me
324 my $size = $self->FETCHSIZE;
325 return if $size == 0;
326 # print STDERR "# POPPITY POP POP POP\n";
327 scalar $self->SPLICE($size-1, 1);
332 scalar $self->SPLICE(0, 1);
337 $self->SPLICE(0, 0, @_);
338 # $self->FETCHSIZE; # av.c takes care of this for me
344 if ($self->{autodefer}) {
345 $self->_annotate_ad_history('CLEAR');
350 $self->{cache}->set_limit($self->{memory});
351 $self->{cache}->empty;
352 @{$self->{offsets}} = (0);
353 %{$self->{deferred}}= ();
354 $self->{deferred_s} = 0;
355 $self->{deferred_max} = -1;
361 # No need to pre-extend anything in this case
362 return if $self->_is_deferring;
364 $self->_fill_offsets_to($n);
365 $self->_extend_file_to($n);
371 if ($self->{autodefer}) {
372 $self->_annotate_ad_history('DELETE');
375 my $lastrec = $self->FETCHSIZE-1;
376 my $rec = $self->FETCH($n);
377 $self->_delete_deferred($n) if $self->_is_deferring;
378 if ($n == $lastrec) {
381 $#{$self->{offsets}}--;
382 $self->{cache}->remove($n);
383 # perhaps in this case I should also remove trailing null records?
385 # Note that delete @a[-3..-1] deletes the records in the wrong order,
386 # so we only chop the very last one out of the file. We could repair this
387 # by tracking deleted records inside the object.
388 } elsif ($n < $lastrec) {
389 $self->STORE($n, "");
396 return 1 if exists $self->{deferred}{$n};
397 $n < $self->FETCHSIZE;
403 if ($self->{autodefer}) {
404 $self->_annotate_ad_history('SPLICE');
407 $self->_flush if $self->_is_deferring; # move this up?
409 $self->_chomp(my @a = $self->_splice(@_));
412 $self->_chomp1(scalar $self->_splice(@_));
418 $self->flush if $self->_is_deferring;
419 $self->{cache}->delink if defined $self->{cache}; # break circular link
420 if ($self->{fh} and $self->{ourfh}) {
421 delete $self->{ourfh};
422 close delete $self->{fh};
427 my ($self, $pos, $nrecs, @data) = @_;
430 $pos = 0 unless defined $pos;
432 # Deal with negative and other out-of-range positions
433 # Also set default for $nrecs
435 my $oldsize = $self->FETCHSIZE;
436 $nrecs = $oldsize unless defined $nrecs;
442 croak "Modification of non-creatable array value attempted, subscript $oldpos";
446 if ($pos > $oldsize) {
448 $pos = $oldsize; # This is what perl does for normal arrays
451 # The manual is very unclear here
453 $nrecs = $oldsize - $pos + $nrecs;
454 $nrecs = 0 if $nrecs < 0;
457 # nrecs is too big---it really means "until the end"
459 if ($nrecs + $pos > $oldsize) {
460 $nrecs = $oldsize - $pos;
464 $self->_fixrecs(@data);
465 my $data = join '', @data;
466 my $datalen = length $data;
469 # compute length of data being removed
470 for ($pos .. $pos+$nrecs-1) {
471 last unless defined $self->_fill_offsets_to($_);
472 my $rec = $self->_fetch($_);
473 last unless defined $rec;
476 # Why don't we just use length($rec) here?
477 # Because that record might have come from the cache. _splice
478 # might have been called to flush out the deferred-write records,
479 # and in this case length($rec) is the length of the record to be
480 # *written*, not the length of the actual record in the file. But
481 # the offsets are still true. 20020322
482 $oldlen += $self->{offsets}[$_+1] - $self->{offsets}[$_]
483 if defined $self->{offsets}[$_+1];
485 $self->_fill_offsets_to($pos+$nrecs);
488 $self->_mtwrite($data, $self->{offsets}[$pos], $oldlen);
489 # Adjust the offsets table
490 $self->_oadjust([$pos, $nrecs, @data]);
492 { # Take this read cache stuff out into a separate function
493 # You made a half-attempt to put it into _oadjust.
494 # Finish something like that up eventually.
495 # STORE also needs to do something similarish
497 # update the read cache, part 1
499 for ($pos .. $pos+$nrecs-1) {
500 my $new = $data[$_-$pos];
502 $self->{cache}->update($_, $new);
504 $self->{cache}->remove($_);
508 # update the read cache, part 2
509 # moved records - records past the site of the change
510 # need to be renumbered
511 # Maybe merge this with the previous block?
513 my @oldkeys = grep $_ >= $pos + $nrecs, $self->{cache}->ckeys;
514 my @newkeys = map $_-$nrecs+@data, @oldkeys;
515 $self->{cache}->rekey(\@oldkeys, \@newkeys);
518 # Now there might be too much data in the cache, if we spliced out
519 # some short records and spliced in some long ones. If so, flush
524 # Yes, the return value of 'splice' *is* actually this complicated
525 wantarray ? @result : @result ? $result[-1] : undef;
529 # write data into the file
530 # $data is the data to be written.
531 # it should be written at position $pos, and should overwrite
532 # exactly $len of the following bytes.
533 # Note that if length($data) > $len, the subsequent bytes will have to
534 # be moved up, and if length($data) < $len, they will have to
537 my ($self, $data, $pos, $len) = @_;
539 unless (defined $pos) {
540 die "\$pos was undefined in _twrite";
543 my $len_diff = length($data) - $len;
545 if ($len_diff == 0) { # Woo-hoo!
546 my $fh = $self->{fh};
548 $self->_write_record($data);
549 return; # well, that was easy.
552 # the two records are of different lengths
553 # our strategy here: rewrite the tail of the file,
554 # reading ahead one buffer at a time
555 # $bufsize is required to be at least as large as the data we're overwriting
556 my $bufsize = _bufsize($len_diff);
557 my ($writepos, $readpos) = ($pos, $pos+$len);
561 # Seems like there ought to be a way to avoid the repeated code
562 # and the special case here. The read(1) is also a little weird.
565 $self->_seekb($readpos);
566 my $br = read $self->{fh}, $next_block, $bufsize;
567 $more_data = read $self->{fh}, my($dummy), 1;
568 $self->_seekb($writepos);
569 $self->_write_record($data);
571 $writepos += length $data;
574 $self->_seekb($writepos);
575 $self->_write_record($next_block);
577 # There might be leftover data at the end of the file
578 $self->_chop_file if $len_diff < 0;
582 # Insert text D at position S.
583 # Let C = E-S-|D|. If C < 0; die.
584 # Data in [S,S+C) is copied to [S+D,S+D+C) = [S+D,E).
585 # Data in [S+C = E-D, E) is returned. Data in [E, oo) is untouched.
587 # In a later version, don't read the entire intervening area into
588 # memory at once; do the copying block by block.
591 my ($D, $s, $e) = @_;
594 local *FH = $self->{fh};
595 confess "Not enough space to insert $d bytes between $s and $e"
597 confess "[$s,$e) is an invalid insertion range" if $e < $s;
600 read FH, my $buf, $e-$s;
602 $D .= substr($buf, 0, $c, "");
605 $self->_write_record($D);
610 # Like _twrite, but the data-pos-len triple may be repeated; you may
611 # write several chunks. All the writing will be done in
612 # one pass. Chunks SHALL be in ascending order and SHALL NOT overlap.
619 or die "Arguments to _mtwrite did not come in groups of three";
622 my ($data, $pos, $len) = splice @_, 0, 3;
623 my $end = $pos + $len; # The OLD end of the segment to be replaced
624 $data = $unwritten . $data;
625 $delta -= length($unwritten);
627 $pos += $delta; # This is where the data goes now
628 my $dlen = length $data;
630 if ($len >= $dlen) { # the data will fit
631 $self->_write_record($data);
632 $delta += ($dlen - $len); # everything following moves down by this much
633 $data = ""; # All the data in the buffer has been written
635 my $writable = substr($data, 0, $len - $delta, "");
636 $self->_write_record($writable);
637 $delta += ($dlen - $len); # everything following moves down by this much
640 # At this point we've written some but maybe not all of the data.
641 # There might be a gap to close up, or $data might still contain a
642 # bunch of unwritten data that didn't fit.
643 my $ndlen = length $data;
645 $self->_write_record($data);
646 } elsif ($delta < 0) {
647 # upcopy (close up gap)
649 $self->_upcopy($end, $end + $delta, $_[1] - $end);
651 $self->_upcopy($end, $end + $delta);
654 # downcopy (insert data that didn't fit; replace this data in memory
655 # with _later_ data that doesn't fit)
657 $unwritten = $self->_downcopy($data, $end, $_[1] - $end);
659 # Make the file longer to accommodate the last segment that doesn'
660 $unwritten = $self->_downcopy($data, $end);
666 # Copy block of data of length $len from position $spos to position $dpos
667 # $dpos must be <= $spos
669 # If $len is undefined, go all the way to the end of the file
670 # and then truncate it ($spos - $dpos bytes will be removed)
672 my $blocksize = 8192;
673 my ($self, $spos, $dpos, $len) = @_;
675 die "source ($spos) was upstream of destination ($dpos) in _upcopy";
676 } elsif ($dpos == $spos) {
680 while (! defined ($len) || $len > 0) {
681 my $readsize = ! defined($len) ? $blocksize
682 : $len > $blocksize ? $blocksize
685 my $fh = $self->{fh};
686 $self->_seekb($spos);
687 my $bytes_read = read $fh, my($data), $readsize;
688 $self->_seekb($dpos);
693 $self->_write_record($data);
694 $spos += $bytes_read;
695 $dpos += $bytes_read;
696 $len -= $bytes_read if defined $len;
700 # Write $data into a block of length $len at position $pos,
701 # moving everything in the block forwards to make room.
702 # Instead of writing the last length($data) bytes from the block
703 # (because there isn't room for them any longer) return them.
705 # Undefined $len means 'until the end of the file'
707 my $blocksize = 8192;
708 my ($self, $data, $pos, $len) = @_;
709 my $fh = $self->{fh};
711 while (! defined $len || $len > 0) {
712 my $readsize = ! defined($len) ? $blocksize
713 : $len > $blocksize? $blocksize : $len;
715 read $fh, my($old), $readsize;
716 my $last_read_was_short = length($old) < $readsize;
719 if ($last_read_was_short) {
720 # If last read was short, then $data now contains the entire rest
721 # of the file, so there's no need to write only one block of it
725 $writable = substr($data, 0, $readsize, "");
727 last if $writable eq "";
729 $self->_write_record($writable);
730 last if $last_read_was_short && $data eq "";
731 $len -= $readsize if defined $len;
737 # Adjust the object data structures following an '_mtwrite'
739 # [$pos, $nrecs, @length] items
740 # indicating that $nrecs records were removed at $recpos (a record offset)
741 # and replaced with records of length @length...
742 # Arguments guarantee that $recpos is strictly increasing.
752 my ($pos, $nrecs, @data) = @$_;
755 # Adjust the offsets of the records after the previous batch up
756 # to the first new one of this batch
757 for my $i ($prev_end+2 .. $pos - 1) {
758 $self->{offsets}[$i] += $delta;
759 $newkey{$i} = $i + $delta_recs;
762 $prev_end = $pos + @data - 1; # last record moved on this pass
764 # Remove the offsets for the removed records;
765 # replace with the offsets for the inserted records
766 my @newoff = ($self->{offsets}[$pos] + $delta);
767 for my $i (0 .. $#data) {
768 my $newlen = length $data[$i];
769 push @newoff, $newoff[$i] + $newlen;
773 for my $i ($pos .. $pos+$nrecs-1) {
774 last if $i+1 > $#{$self->{offsets}};
775 my $oldlen = $self->{offsets}[$i+1] - $self->{offsets}[$i];
779 # # also this data has changed, so update it in the cache
780 # for (0 .. $#data) {
781 # $self->{cache}->update($pos + $_, $data[$_]);
784 # my @oldkeys = grep $_ >= $pos + @data, $self->{cache}->ckeys;
785 # my @newkeys = map $_ + $delta_recs, @oldkeys;
786 # $self->{cache}->rekey(\@oldkeys, \@newkeys);
789 # replace old offsets with new
790 splice @{$self->{offsets}}, $pos, $nrecs+1, @newoff;
791 # What if we just spliced out the end of the offsets table?
792 # shouldn't we clear $self->{eof}? Test for this XXX BUG TODO
794 $delta_recs += @data - $nrecs; # net change in total number of records
797 # The trailing records at the very end of the file
799 for my $i ($prev_end+2 .. $#{$self->{offsets}}) {
800 $self->{offsets}[$i] += $delta;
804 # If we scrubbed out all known offsets, regenerate the trivial table
805 # that knows that the file does indeed start at 0.
806 $self->{offsets}[0] = 0 unless @{$self->{offsets}};
807 # If the file got longer, the offsets table is no longer complete
808 # $self->{eof} = 0 if $delta_recs > 0;
810 # Now there might be too much data in the cache, if we spliced out
811 # some short records and spliced in some long ones. If so, flush
816 # If a record does not already end with the appropriate terminator
817 # string, append one.
821 $_ = "" unless defined $_;
822 $_ .= $self->{recsep}
823 unless substr($_, - $self->{recseplen}) eq $self->{recsep};
828 ################################################################
830 # Basic read, write, and seek
833 # seek to the beginning of record #$n
834 # Assumes that the offsets table is already correctly populated
836 # Note that $n=-1 has a special meaning here: It means the start of
837 # the last known record; this may or may not be the very last record
838 # in the file, depending on whether the offsets table is fully populated.
842 my $o = $self->{offsets}[$n];
844 or confess("logic error: undefined offset for record $n");
845 seek $self->{fh}, $o, SEEK_SET
846 or confess "Couldn't seek filehandle: $!"; # "Should never happen."
849 # seek to byte $b in the file
852 seek $self->{fh}, $b, SEEK_SET
853 or die "Couldn't seek filehandle: $!"; # "Should never happen."
856 # populate the offsets table up to the beginning of record $n
857 # return the offset of record $n
858 sub _fill_offsets_to {
861 return $self->{offsets}[$n] if $self->{eof};
863 my $fh = $self->{fh};
864 local *OFF = $self->{offsets};
867 until ($#OFF >= $n) {
868 $self->_seek(-1); # tricky -- see comment at _seek
869 $rec = $self->_read_record;
871 push @OFF, int(tell $fh); # Tels says that int() saves memory here
874 return; # It turns out there is no such record
878 # we have now read all the records up to record n-1,
879 # so we can return the offset of record n
886 my $fh = $self->{fh};
887 local *OFF = $self->{offsets};
889 $self->_seek(-1); # tricky -- see comment at _seek
891 # Tels says that inlining read_record() would make this loop
892 # five times faster. 20030508
893 while ( defined $self->_read_record()) {
894 # int() saves us memory here
895 push @OFF, int(tell $fh);
902 # assumes that $rec is already suitably terminated
904 my ($self, $rec) = @_;
905 my $fh = $self->{fh};
908 or die "Couldn't write record: $!"; # "Should never happen."
909 # $self->{_written} += length($rec);
915 { local $/ = $self->{recsep};
916 my $fh = $self->{fh};
919 return unless defined $rec;
920 if (substr($rec, -$self->{recseplen}) ne $self->{recsep}) {
921 # improperly terminated final record --- quietly fix it.
922 # my $ac = substr($rec, -$self->{recseplen});
924 $self->{sawlastrec} = 1;
925 unless ($self->{rdonly}) {
927 my $fh = $self->{fh};
928 print $fh $self->{recsep};
930 $rec .= $self->{recsep};
932 # $self->{_read} += length($rec) if defined $rec;
938 @{$self}{'_read', '_written'};
941 ################################################################
943 # Read cache management
947 $self->{cache}->reduce_size_to($self->{memory} - $self->{deferred_s});
950 sub _cache_too_full {
952 $self->{cache}->bytes + $self->{deferred_s} >= $self->{memory};
955 ################################################################
957 # File custodial services
961 # We have read to the end of the file and have the offsets table
962 # entirely populated. Now we need to write a new record beyond
963 # the end of the file. We prepare for this by writing
964 # empty records into the file up to the position we want
966 # assumes that the offsets table already contains the offset of record $n,
967 # if it exists, and extends to the end of the file if not.
968 sub _extend_file_to {
970 $self->_seek(-1); # position after the end of the last record
971 my $pos = $self->{offsets}[-1];
973 # the offsets table has one entry more than the total number of records
974 my $extras = $n - $#{$self->{offsets}};
976 # Todo : just use $self->{recsep} x $extras here?
977 while ($extras-- > 0) {
978 $self->_write_record($self->{recsep});
979 push @{$self->{offsets}}, int(tell $self->{fh});
983 # Truncate the file at the current position
986 truncate $self->{fh}, tell($self->{fh});
990 # compute the size of a buffer suitable for moving
991 # all the data in a file forward $n bytes
992 # ($n may be negative)
993 # The result should be at least $n.
996 return 8192 if $n <= 0;
998 $b += 8192 if $n & 8191;
1002 ################################################################
1004 # Miscellaneous public methods
1009 my ($self, $op) = @_;
1011 my $pack = ref $self;
1012 croak "Usage: $pack\->flock([OPERATION])";
1014 my $fh = $self->{fh};
1015 $op = LOCK_EX unless defined $op;
1016 my $locked = flock $fh, $op;
1018 if ($locked && ($op & (LOCK_EX | LOCK_SH))) {
1019 # If you're locking the file, then presumably it's because
1020 # there might have been a write access by another process.
1021 # In that case, the read cache contents and the offsets table
1022 # might be invalid, so discard them. 20030508
1023 $self->{offsets} = [0];
1024 $self->{cache}->empty;
1030 # Get/set autochomp option
1034 my $old = $self->{autochomp};
1035 $self->{autochomp} = shift;
1042 # Get offset table entries; returns offset of nth record
1044 my ($self, $n) = @_;
1046 if ($#{$self->{offsets}} < $n) {
1047 return if $self->{eof}; # request for record beyond the end of file
1048 my $o = $self->_fill_offsets_to($n);
1049 # If it's still undefined, there is no such record, so return 'undef'
1050 return unless defined $o;
1053 $self->{offsets}[$n];
1056 sub discard_offsets {
1058 $self->{offsets} = [0];
1061 ################################################################
1063 # Matters related to deferred writing
1069 $self->_stop_autodeferring;
1070 @{$self->{ad_history}} = ();
1074 # Flush deferred writes
1076 # This could be better optimized to write the file in one pass, instead
1077 # of one pass per block of records. But that will require modifications
1078 # to _twrite, so I should have a good _twrite test suite first.
1088 my @writable = sort {$a<=>$b} (keys %{$self->{deferred}});
1091 # gather all consecutive records from the front of @writable
1092 my $first_rec = shift @writable;
1093 my $last_rec = $first_rec+1;
1094 ++$last_rec, shift @writable while @writable && $last_rec == $writable[0];
1096 $self->_fill_offsets_to($last_rec);
1097 $self->_extend_file_to($last_rec);
1098 $self->_splice($first_rec, $last_rec-$first_rec+1,
1099 @{$self->{deferred}}{$first_rec .. $last_rec});
1102 $self->_discard; # clear out defered-write-cache
1107 my @writable = sort {$a<=>$b} (keys %{$self->{deferred}});
1112 # gather all consecutive records from the front of @writable
1113 my $first_rec = shift @writable;
1114 my $last_rec = $first_rec+1;
1115 ++$last_rec, shift @writable while @writable && $last_rec == $writable[0];
1117 my $end = $self->_fill_offsets_to($last_rec+1);
1118 if (not defined $end) {
1119 $self->_extend_file_to($last_rec);
1120 $end = $self->{offsets}[$last_rec];
1122 my ($start) = $self->{offsets}[$first_rec];
1124 join("", @{$self->{deferred}}{$first_rec .. $last_rec}), # data
1126 $end-$start; # length
1127 push @adjust, [$first_rec, # starting at this position...
1128 $last_rec-$first_rec+1, # this many records...
1129 # are replaced with these...
1130 @{$self->{deferred}}{$first_rec .. $last_rec},
1134 $self->_mtwrite(@args); # write multiple record groups
1135 $self->_discard; # clear out defered-write-cache
1136 $self->_oadjust(@adjust);
1139 # Discard deferred writes and disable future deferred writes
1146 # Discard deferred writes, but retain old deferred writing mode
1149 %{$self->{deferred}} = ();
1150 $self->{deferred_s} = 0;
1151 $self->{deferred_max} = -1;
1152 $self->{cache}->set_limit($self->{memory});
1155 # Deferred writing is enabled, either explicitly ($self->{defer})
1156 # or automatically ($self->{autodeferring})
1159 $self->{defer} || $self->{autodeferring};
1162 # The largest record number of any deferred record
1165 return $self->{deferred_max} if defined $self->{deferred_max};
1167 for my $key (keys %{$self->{deferred}}) {
1168 $max = $key if $key > $max;
1170 $self->{deferred_max} = $max;
1174 ################################################################
1176 # Matters related to autodeferment
1179 # Get/set autodefer option
1183 my $old = $self->{autodefer};
1184 $self->{autodefer} = shift;
1186 $self->_stop_autodeferring;
1187 @{$self->{ad_history}} = ();
1195 # The user is trying to store record #$n Record that in the history,
1196 # and then enable (or disable) autodeferment if that seems useful.
1197 # Note that it's OK for $n to be a non-number, as long as the function
1198 # is prepared to deal with that. Nobody else looks at the ad_history.
1200 # Now, what does the ad_history mean, and what is this function doing?
1201 # Essentially, the idea is to enable autodeferring when we see that the
1202 # user has made three consecutive STORE calls to three consecutive records.
1203 # ("Three" is actually ->{autodefer_threshhold}.)
1204 # A STORE call for record #$n inserts $n into the autodefer history,
1205 # and if the history contains three consecutive records, we enable
1206 # autodeferment. An ad_history of [X, Y] means that the most recent
1207 # STOREs were for records X, X+1, ..., Y, in that order.
1209 # Inserting a nonconsecutive number erases the history and starts over.
1211 # Performing a special operation like SPLICE erases the history.
1213 # There's one special case: CLEAR means that CLEAR was just called.
1214 # In this case, we prime the history with [-2, -1] so that if the next
1215 # write is for record 0, autodeferring goes on immediately. This is for
1216 # the common special case of "@a = (...)".
1218 sub _annotate_ad_history {
1219 my ($self, $n) = @_;
1220 return unless $self->{autodefer}; # feature is disabled
1221 return if $self->{defer}; # already in explicit defer mode
1222 return unless $self->{offsets}[-1] >= $self->{autodefer_filelen_threshhold};
1224 local *H = $self->{ad_history};
1225 if ($n eq 'CLEAR') {
1226 @H = (-2, -1); # prime the history with fake records
1227 $self->_stop_autodeferring;
1228 } elsif ($n =~ /^\d+$/) {
1232 if ($H[1] == $n-1) { # another consecutive record
1234 if ($H[1] - $H[0] + 1 >= $self->{autodefer_threshhold}) {
1235 $self->{autodeferring} = 1;
1237 } else { # nonconsecutive- erase and start over
1239 $self->_stop_autodeferring;
1242 } else { # SPLICE or STORESIZE or some such
1244 $self->_stop_autodeferring;
1248 # If autodeferring was enabled, cut it out and discard the history
1249 sub _stop_autodeferring {
1251 if ($self->{autodeferring}) {
1254 $self->{autodeferring} = 0;
1257 ################################################################
1260 # This is NOT a method. It is here for two reasons:
1261 # 1. To factor a fairly complicated block out of the constructor
1262 # 2. To provide access for the test suite, which need to be sure
1263 # files are being written properly.
1264 sub _default_recsep {
1266 if ($^O eq 'MSWin32') { # Dos too?
1267 # Windows users expect files to be terminated with \r\n
1268 # But $/ is set to \n instead
1269 # Note that this also transforms \n\n into \r\n\r\n.
1270 # That is a feature.
1271 $recsep =~ s/\n/\r\n/g;
1276 # Utility function for _check_integrity
1284 # Given a file, make sure the cache is consistent with the
1285 # file contents and the internal data structures are consistent with
1286 # each other. Returns true if everything checks out, false if not
1288 # The $file argument is no longer used. It is retained for compatibility
1289 # with the existing test suite.
1290 sub _check_integrity {
1291 my ($self, $file, $warn) = @_;
1292 my $rsl = $self->{recseplen};
1293 my $rs = $self->{recsep};
1295 local *_; # local $_ does not work here
1296 local $DIAGNOSTIC = 1;
1298 if (not defined $rs) {
1299 _ci_warn("recsep is undef!");
1301 } elsif ($rs eq "") {
1302 _ci_warn("recsep is empty!");
1304 } elsif ($rsl != length $rs) {
1305 my $ln = length $rs;
1306 _ci_warn("recsep <$rs> has length $ln, should be $rsl");
1310 if (not defined $self->{offsets}[0]) {
1311 _ci_warn("offset 0 is missing!");
1314 } elsif ($self->{offsets}[0] != 0) {
1315 _ci_warn("rec 0: offset <$self->{offsets}[0]> s/b 0!");
1321 local *F = $self->{fh};
1322 seek F, 0, SEEK_SET;
1328 my $cached = $self->{cache}->_produce($n);
1329 my $offset = $self->{offsets}[$.];
1331 if (defined $offset && $offset != $ao) {
1332 _ci_warn("rec $n: offset <$offset> actual <$ao>");
1335 if (defined $cached && $_ ne $cached && ! $self->{deferred}{$n}) {
1337 _ci_warn("rec $n: cached <$cached> actual <$_>");
1339 if (defined $cached && substr($cached, -$rsl) ne $rs) {
1341 _ci_warn("rec $n in the cache is missing the record separator");
1343 if (! defined $offset && $self->{eof}) {
1345 _ci_warn("The offset table was marked complete, but it is missing element $.");
1348 if (@{$self->{offsets}} > $.+1) {
1350 my $n = @{$self->{offsets}};
1351 _ci_warn("The offset table has $n items, but the file has only $.");
1354 my $deferring = $self->_is_deferring;
1355 for my $n ($self->{cache}->ckeys) {
1356 my $r = $self->{cache}->_produce($n);
1357 $cached += length($r);
1358 next if $n+1 <= $.; # checked this already
1359 _ci_warn("spurious caching of record $n");
1362 my $b = $self->{cache}->bytes;
1363 if ($cached != $b) {
1364 _ci_warn("cache size is $b, should be $cached");
1369 # That cache has its own set of tests
1370 $good = 0 unless $self->{cache}->_check_integrity;
1372 # Now let's check the deferbuffer
1373 # Unless deferred writing is enabled, it should be empty
1374 if (! $self->_is_deferring && %{$self->{deferred}}) {
1375 _ci_warn("deferred writing disabled, but deferbuffer nonempty");
1379 # Any record in the deferbuffer should *not* be present in the readcache
1381 while (my ($n, $r) = each %{$self->{deferred}}) {
1382 $deferred_s += length($r);
1383 if (defined $self->{cache}->_produce($n)) {
1384 _ci_warn("record $n is in the deferbuffer *and* the readcache");
1387 if (substr($r, -$rsl) ne $rs) {
1388 _ci_warn("rec $n in the deferbuffer is missing the record separator");
1393 # Total size of deferbuffer should match internal total
1394 if ($deferred_s != $self->{deferred_s}) {
1395 _ci_warn("buffer size is $self->{deferred_s}, should be $deferred_s");
1399 # Total size of deferbuffer should not exceed the specified limit
1400 if ($deferred_s > $self->{dw_size}) {
1401 _ci_warn("buffer size is $self->{deferred_s} which exceeds the limit of $self->{dw_size}");
1405 # Total size of cached data should not exceed the specified limit
1406 if ($deferred_s + $cached > $self->{memory}) {
1407 my $total = $deferred_s + $cached;
1408 _ci_warn("total stored data size is $total which exceeds the limit of $self->{memory}");
1412 # Stuff related to autodeferment
1413 if (!$self->{autodefer} && @{$self->{ad_history}}) {
1414 _ci_warn("autodefer is disabled, but ad_history is nonempty");
1417 if ($self->{autodeferring} && $self->{defer}) {
1418 _ci_warn("both autodeferring and explicit deferring are active");
1421 if (@{$self->{ad_history}} == 0) {
1422 # That's OK, no additional tests required
1423 } elsif (@{$self->{ad_history}} == 2) {
1424 my @non_number = grep !/^-?\d+$/, @{$self->{ad_history}};
1428 $msg = "ad_history contains non-numbers (@{$self->{ad_history}})";
1432 } elsif ($self->{ad_history}[1] < $self->{ad_history}[0]) {
1433 _ci_warn("ad_history has nonsensical values @{$self->{ad_history}}");
1437 _ci_warn("ad_history has bad length <@{$self->{ad_history}}>");
1444 ################################################################
1450 package Tie::File::Cache;
1451 $Tie::File::Cache::VERSION = $Tie::File::VERSION;
1452 use Carp ':DEFAULT', 'confess';
1458 #sub STAT () { 4 } # Array with request statistics for each record
1459 #sub MISS () { 5 } # Total number of cache misses
1460 #sub REQ () { 6 } # Total number of cache requests
1464 my ($pack, $max) = @_;
1466 croak "missing argument to ->new" unless defined $max;
1468 bless $self => $pack;
1469 @$self = (Tie::File::Heap->new($self), {}, $max, 0);
1474 my ($self, $n) = @_;
1479 my ($self, $n) = @_;
1483 # For internal use only
1484 # Will be called by the heap structure to notify us that a certain
1485 # piece of data has moved from one heap element to another.
1486 # $k is the hash key of the item
1487 # $n is the new index into the heap at which it is stored
1488 # If $n is undefined, the item has been removed from the heap.
1490 my ($self, $k, $n) = @_;
1492 $self->[HASH]{$k} = $n;
1494 delete $self->[HASH]{$k};
1499 my ($self, $key, $val) = @_;
1501 croak "missing argument to ->insert" unless defined $key;
1502 unless (defined $self->[MAX]) {
1503 confess "undefined max" ;
1505 confess "undefined val" unless defined $val;
1506 return if length($val) > $self->[MAX];
1508 # if ($self->[STAT]) {
1509 # $self->[STAT][$key] = 1;
1513 my $oldnode = $self->[HASH]{$key};
1514 if (defined $oldnode) {
1515 my $oldval = $self->[HEAP]->set_val($oldnode, $val);
1516 $self->[BYTES] -= length($oldval);
1518 $self->[HEAP]->insert($key, $val);
1520 $self->[BYTES] += length($val);
1521 $self->flush if $self->[BYTES] > $self->[MAX];
1526 my $old_data = $self->[HEAP]->popheap;
1527 return unless defined $old_data;
1528 $self->[BYTES] -= length $old_data;
1533 my ($self, @keys) = @_;
1536 # if ($self->[STAT]) {
1537 # for my $key (@keys) {
1538 # $self->[STAT][$key] = 0;
1543 for my $key (@keys) {
1544 next unless exists $self->[HASH]{$key};
1545 my $old_data = $self->[HEAP]->remove($self->[HASH]{$key});
1546 $self->[BYTES] -= length $old_data;
1547 push @result, $old_data;
1553 my ($self, $key) = @_;
1555 croak "missing argument to ->lookup" unless defined $key;
1557 # if ($self->[STAT]) {
1558 # $self->[MISS]++ if $self->[STAT][$key]++ == 0;
1560 # my $hit_rate = 1 - $self->[MISS] / $self->[REQ];
1561 # # Do some testing to determine this threshhold
1562 # $#$self = STAT - 1 if $hit_rate > 0.20;
1565 if (exists $self->[HASH]{$key}) {
1566 $self->[HEAP]->lookup($self->[HASH]{$key});
1572 # For internal use only
1574 my ($self, $key) = @_;
1575 my $loc = $self->[HASH]{$key};
1576 return unless defined $loc;
1577 $self->[HEAP][$loc][2];
1580 # For internal use only
1582 my ($self, $key) = @_;
1583 $self->[HEAP]->promote($self->[HASH]{$key});
1588 %{$self->[HASH]} = ();
1590 $self->[HEAP]->empty;
1591 # @{$self->[STAT]} = ();
1592 # $self->[MISS] = 0;
1598 keys %{$self->[HASH]} == 0;
1602 my ($self, $key, $val) = @_;
1604 croak "missing argument to ->update" unless defined $key;
1605 if (length($val) > $self->[MAX]) {
1606 my ($oldval) = $self->remove($key);
1607 $self->[BYTES] -= length($oldval) if defined $oldval;
1608 } elsif (exists $self->[HASH]{$key}) {
1609 my $oldval = $self->[HEAP]->set_val($self->[HASH]{$key}, $val);
1610 $self->[BYTES] += length($val);
1611 $self->[BYTES] -= length($oldval) if defined $oldval;
1613 $self->[HEAP]->insert($key, $val);
1614 $self->[BYTES] += length($val);
1620 my ($self, $okeys, $nkeys) = @_;
1623 @map{@$okeys} = @$nkeys;
1624 croak "missing argument to ->rekey" unless defined $nkeys;
1625 croak "length mismatch in ->rekey arguments" unless @$nkeys == @$okeys;
1626 my %adjusted; # map new keys to heap indices
1627 # You should be able to cut this to one loop TODO XXX
1628 for (0 .. $#$okeys) {
1629 $adjusted{$nkeys->[$_]} = delete $self->[HASH]{$okeys->[$_]};
1631 while (my ($nk, $ix) = each %adjusted) {
1632 # @{$self->[HASH]}{keys %adjusted} = values %adjusted;
1633 $self->[HEAP]->rekey($ix, $nk);
1634 $self->[HASH]{$nk} = $ix;
1640 my @a = keys %{$self->[HASH]};
1644 # Return total amount of cached data
1650 # Expire oldest item from cache until cache size is smaller than $max
1651 sub reduce_size_to {
1652 my ($self, $max) = @_;
1653 until ($self->[BYTES] <= $max) {
1654 # Note that Tie::File::Cache::expire has been inlined here
1655 my $old_data = $self->[HEAP]->popheap;
1656 return unless defined $old_data;
1657 $self->[BYTES] -= length $old_data;
1661 # Why not just $self->reduce_size_to($self->[MAX])?
1662 # Try this when things stabilize TODO XXX
1663 # If the cache is too full, expire the oldest records
1666 $self->reduce_size_to($self->[MAX]) if $self->[BYTES] > $self->[MAX];
1669 # For internal use only
1672 $self->[HEAP]->expire_order;
1675 BEGIN { *_ci_warn = \&Tie::File::_ci_warn }
1677 sub _check_integrity { # For CACHE
1682 $self->[HEAP]->_check_integrity or $good = 0;
1686 for my $k (keys %{$self->[HASH]}) {
1687 if ($k ne '0' && $k !~ /^[1-9][0-9]*$/) {
1689 _ci_warn "Cache hash key <$k> is non-numeric";
1692 my $h = $self->[HASH]{$k};
1695 _ci_warn "Heap index number for key $k is undefined";
1698 _ci_warn "Heap index number for key $k is zero";
1700 my $j = $self->[HEAP][$h];
1703 _ci_warn "Heap contents key $k (=> $h) are undefined";
1705 $bytes += length($j->[2]);
1706 if ($k ne $j->[1]) {
1708 _ci_warn "Heap contents key $k (=> $h) is $j->[1], should be $k";
1715 if ($bytes != $self->[BYTES]) {
1717 _ci_warn "Total data in cache is $bytes, expected $self->[BYTES]";
1721 if ($bytes > $self->[MAX]) {
1723 _ci_warn "Total data in cache is $bytes, exceeds maximum $self->[MAX]";
1731 $self->[HEAP] = undef; # Bye bye heap
1734 ################################################################
1738 # Heap data structure for use by cache LRU routines
1740 package Tie::File::Heap;
1741 use Carp ':DEFAULT', 'confess';
1742 $Tie::File::Heap::VERSION = $Tie::File::Cache::VERSION;
1748 my ($pack, $cache) = @_;
1749 die "$pack: Parent cache object $cache does not support _heap_move method"
1750 unless eval { $cache->can('_heap_move') };
1751 my $self = [[0,$cache,0]];
1752 bless $self => $pack;
1755 # Allocate a new sequence number, larger than all previously allocated numbers
1790 $self->[0][0] = 0; # might as well reset the sequence numbers
1793 # notify the parent cache object that we moved something
1796 $self->_cache->_heap_move(@_);
1799 # Insert a piece of data into the heap with the indicated sequence number.
1800 # The item with the smallest sequence number is always at the top.
1801 # If no sequence number is specified, allocate a new one and insert the
1802 # item at the bottom.
1804 my ($self, $key, $data, $seq) = @_;
1805 $seq = $self->_nseq unless defined $seq;
1806 $self->_insert_new([$seq, $key, $data]);
1809 # Insert a new, fresh item at the bottom of the heap
1811 my ($self, $item) = @_;
1813 $i = int($i/2) until defined $self->[$i/2];
1814 $self->[$i] = $item;
1815 $self->[0][1]->_heap_move($self->[$i][KEY], $i);
1819 # Insert [$data, $seq] pair at or below item $i in the heap.
1820 # If $i is omitted, default to 1 (the top element.)
1822 my ($self, $item, $i) = @_;
1823 # $self->_check_loc($i) if defined $i;
1824 $i = 1 unless defined $i;
1825 until (! defined $self->[$i]) {
1826 if ($self->[$i][SEQ] > $item->[SEQ]) { # inserted item is older
1827 ($self->[$i], $item) = ($item, $self->[$i]);
1828 $self->[0][1]->_heap_move($self->[$i][KEY], $i);
1830 # If either is undefined, go that way. Otherwise, choose at random
1832 $dir = 0 if !defined $self->[2*$i];
1833 $dir = 1 if !defined $self->[2*$i+1];
1834 $dir = int(rand(2)) unless defined $dir;
1837 $self->[$i] = $item;
1838 $self->[0][1]->_heap_move($self->[$i][KEY], $i);
1842 # Remove the item at node $i from the heap, moving child items upwards.
1843 # The item with the smallest sequence number is always at the top.
1844 # Moving items upwards maintains this condition.
1845 # Return the removed item. Return undef if there was no item at node $i.
1847 my ($self, $i) = @_;
1848 $i = 1 unless defined $i;
1849 my $top = $self->[$i];
1850 return unless defined $top;
1853 my ($L, $R) = (2*$i, 2*$i+1);
1855 # If either is undefined, go the other way.
1856 # Otherwise, go towards the smallest.
1857 last unless defined $self->[$L] || defined $self->[$R];
1858 $ii = $R if not defined $self->[$L];
1859 $ii = $L if not defined $self->[$R];
1860 unless (defined $ii) {
1861 $ii = $self->[$L][SEQ] < $self->[$R][SEQ] ? $L : $R;
1864 $self->[$i] = $self->[$ii]; # Promote child to fill vacated spot
1865 $self->[0][1]->_heap_move($self->[$i][KEY], $i);
1866 $i = $ii; # Fill new vacated spot
1868 $self->[0][1]->_heap_move($top->[KEY], undef);
1879 # set the sequence number of the indicated item to a higher number
1880 # than any other item in the heap, and bubble the item down to the
1883 my ($self, $n) = @_;
1884 # $self->_check_loc($n);
1885 $self->[$n][SEQ] = $self->_nseq;
1888 my ($L, $R) = (2*$i, 2*$i+1);
1890 last unless defined $self->[$L] || defined $self->[$R];
1891 $dir = $R unless defined $self->[$L];
1892 $dir = $L unless defined $self->[$R];
1893 unless (defined $dir) {
1894 $dir = $self->[$L][SEQ] < $self->[$R][SEQ] ? $L : $R;
1896 @{$self}[$i, $dir] = @{$self}[$dir, $i];
1898 $self->[0][1]->_heap_move($self->[$_][KEY], $_) if defined $self->[$_];
1904 # Return item $n from the heap, promoting its LRU status
1906 my ($self, $n) = @_;
1907 # $self->_check_loc($n);
1908 my $val = $self->[$n];
1914 # Assign a new value for node $n, promoting it to the bottom of the heap
1916 my ($self, $n, $val) = @_;
1917 # $self->_check_loc($n);
1918 my $oval = $self->[$n][DAT];
1919 $self->[$n][DAT] = $val;
1924 # The hask key has changed for an item;
1925 # alter the heap's record of the hash key
1927 my ($self, $n, $new_key) = @_;
1928 # $self->_check_loc($n);
1929 $self->[$n][KEY] = $new_key;
1933 my ($self, $n) = @_;
1934 unless (1 || defined $self->[$n]) {
1935 confess "_check_loc($n) failed";
1939 BEGIN { *_ci_warn = \&Tie::File::_ci_warn }
1941 sub _check_integrity {
1946 unless (eval {$self->[0][1]->isa("Tie::File::Cache")}) {
1947 _ci_warn "Element 0 of heap corrupt";
1950 $good = 0 unless $self->_satisfies_heap_condition(1);
1951 for my $i (2 .. $#{$self}) {
1952 my $p = int($i/2); # index of parent node
1953 if (defined $self->[$i] && ! defined $self->[$p]) {
1954 _ci_warn "Element $i of heap defined, but parent $p isn't";
1958 if (defined $self->[$i]) {
1959 if ($seq{$self->[$i][SEQ]}) {
1960 my $seq = $self->[$i][SEQ];
1961 _ci_warn "Nodes $i and $seq{$seq} both have SEQ=$seq";
1964 $seq{$self->[$i][SEQ]} = $i;
1972 sub _satisfies_heap_condition {
1978 next unless defined $self->[$c];
1979 if ($self->[$n][SEQ] >= $self->[$c]) {
1980 _ci_warn "Node $n of heap does not predate node $c";
1983 $good = 0 unless $self->_satisfies_heap_condition($c);
1988 # Return a list of all the values, sorted by expiration order
1991 my @nodes = sort {$a->[SEQ] <=> $b->[SEQ]} $self->_nodes;
1992 map { $_->[KEY] } @nodes;
1998 return unless defined $self->[$i];
1999 ($self->[$i], $self->_nodes($i*2), $self->_nodes($i*2+1));
2002 "Cogito, ergo sum."; # don't forget to return a true value from the file
2008 Tie::File - Access the lines of a disk file via a Perl array
2012 # This file documents Tie::File version 0.97
2015 tie @array, 'Tie::File', filename or die ...;
2017 $array[13] = 'blah'; # line 13 of the file is now 'blah'
2018 print $array[42]; # display line 42 of the file
2020 $n_recs = @array; # how many records are in the file?
2021 $#array -= 2; # chop two records off the end
2025 s/PERL/Perl/g; # Replace PERL with Perl everywhere in the file
2028 # These are just like regular push, pop, unshift, shift, and splice
2029 # Except that they modify the file in the way you would expect
2031 push @array, new recs...;
2032 my $r1 = pop @array;
2033 unshift @array, new recs...;
2034 my $r2 = shift @array;
2035 @old_recs = splice @array, 3, 7, new recs...;
2037 untie @array; # all finished
2042 C<Tie::File> represents a regular text file as a Perl array. Each
2043 element in the array corresponds to a record in the file. The first
2044 line of the file is element 0 of the array; the second line is element
2047 The file is I<not> loaded into memory, so this will work even for
2050 Changes to the array are reflected in the file immediately.
2052 Lazy people and beginners may now stop reading the manual.
2056 What is a 'record'? By default, the meaning is the same as for the
2057 C<E<lt>...E<gt>> operator: It's a string terminated by C<$/>, which is
2058 probably C<"\n">. (Minor exception: on DOS and Win32 systems, a
2059 'record' is a string terminated by C<"\r\n">.) You may change the
2060 definition of "record" by supplying the C<recsep> option in the C<tie>
2063 tie @array, 'Tie::File', $file, recsep => 'es';
2065 This says that records are delimited by the string C<es>. If the file
2066 contained the following data:
2068 Curse these pesky flies!\n
2070 then the C<@array> would appear to have four elements:
2077 An undefined value is not permitted as a record separator. Perl's
2078 special "paragraph mode" semantics (E<agrave> la C<$/ = "">) are not
2081 Records read from the tied array do not have the record separator
2082 string on the end; this is to allow
2084 $array[17] .= "extra";
2086 to work as expected.
2088 (See L<"autochomp">, below.) Records stored into the array will have
2089 the record separator string appended before they are written to the
2090 file, if they don't have one already. For example, if the record
2091 separator string is C<"\n">, then the following two lines do exactly
2094 $array[17] = "Cherry pie";
2095 $array[17] = "Cherry pie\n";
2097 The result is that the contents of line 17 of the file will be
2098 replaced with "Cherry pie"; a newline character will separate line 17
2099 from line 18. This means that this code will do nothing:
2103 Because the C<chomp>ed value will have the separator reattached when
2104 it is written back to the file. There is no way to create a file
2105 whose trailing record separator string is missing.
2107 Inserting records that I<contain> the record separator string is not
2108 supported by this module. It will probably produce a reasonable
2109 result, but what this result will be may change in a future version.
2110 Use 'splice' to insert records or to replace one record with several.
2114 Normally, array elements have the record separator removed, so that if
2115 the file contains the text
2121 the tied array will appear to contain C<("Gold", "Frankincense",
2122 "Myrrh")>. If you set C<autochomp> to a false value, the record
2123 separator will not be removed. If the file above was tied with
2125 tie @gifts, "Tie::File", $gifts, autochomp => 0;
2127 then the array C<@gifts> would appear to contain C<("Gold\n",
2128 "Frankincense\n", "Myrrh\n")>, or (on Win32 systems) C<("Gold\r\n",
2129 "Frankincense\r\n", "Myrrh\r\n")>.
2133 Normally, the specified file will be opened for read and write access,
2134 and will be created if it does not exist. (That is, the flags
2135 C<O_RDWR | O_CREAT> are supplied in the C<open> call.) If you want to
2136 change this, you may supply alternative flags in the C<mode> option.
2137 See L<Fcntl> for a listing of available flags.
2140 # open the file if it exists, but fail if it does not exist
2142 tie @array, 'Tie::File', $file, mode => O_RDWR;
2144 # create the file if it does not exist
2145 use Fcntl 'O_RDWR', 'O_CREAT';
2146 tie @array, 'Tie::File', $file, mode => O_RDWR | O_CREAT;
2148 # open an existing file in read-only mode
2149 use Fcntl 'O_RDONLY';
2150 tie @array, 'Tie::File', $file, mode => O_RDONLY;
2152 Opening the data file in write-only or append mode is not supported.
2156 This is an upper limit on the amount of memory that C<Tie::File> will
2157 consume at any time while managing the file. This is used for two
2158 things: managing the I<read cache> and managing the I<deferred write
2161 Records read in from the file are cached, to avoid having to re-read
2162 them repeatedly. If you read the same record twice, the first time it
2163 will be stored in memory, and the second time it will be fetched from
2164 the I<read cache>. The amount of data in the read cache will not
2165 exceed the value you specified for C<memory>. If C<Tie::File> wants
2166 to cache a new record, but the read cache is full, it will make room
2167 by expiring the least-recently visited records from the read cache.
2169 The default memory limit is 2Mib. You can adjust the maximum read
2170 cache size by supplying the C<memory> option. The argument is the
2171 desired cache size, in bytes.
2173 # I have a lot of memory, so use a large cache to speed up access
2174 tie @array, 'Tie::File', $file, memory => 20_000_000;
2176 Setting the memory limit to 0 will inhibit caching; records will be
2177 fetched from disk every time you examine them.
2179 The C<memory> value is not an absolute or exact limit on the memory
2180 used. C<Tie::File> objects contains some structures besides the read
2181 cache and the deferred write buffer, whose sizes are not charged
2184 The cache itself consumes about 310 bytes per cached record, so if
2185 your file has many short records, you may want to decrease the cache
2186 memory limit, or else the cache overhead may exceed the size of the
2192 (This is an advanced feature. Skip this section on first reading.)
2194 If you use deferred writing (See L<"Deferred Writing">, below) then
2195 data you write into the array will not be written directly to the
2196 file; instead, it will be saved in the I<deferred write buffer> to be
2197 written out later. Data in the deferred write buffer is also charged
2198 against the memory limit you set with the C<memory> option.
2200 You may set the C<dw_size> option to limit the amount of data that can
2201 be saved in the deferred write buffer. This limit may not exceed the
2202 total memory limit. For example, if you set C<dw_size> to 1000 and
2203 C<memory> to 2500, that means that no more than 1000 bytes of deferred
2204 writes will be saved up. The space available for the read cache will
2205 vary, but it will always be at least 1500 bytes (if the deferred write
2206 buffer is full) and it could grow as large as 2500 bytes (if the
2207 deferred write buffer is empty.)
2209 If you don't specify a C<dw_size>, it defaults to the entire memory
2212 =head2 Option Format
2214 C<-mode> is a synonym for C<mode>. C<-recsep> is a synonym for
2215 C<recsep>. C<-memory> is a synonym for C<memory>. You get the
2218 =head1 Public Methods
2220 The C<tie> call returns an object, say C<$o>. You may call
2222 $rec = $o->FETCH($n);
2223 $o->STORE($n, $rec);
2225 to fetch or store the record at line C<$n>, respectively; similarly
2226 the other tied array methods. (See L<perltie> for details.) You may
2227 also call the following methods on this object:
2233 will lock the tied file. C<MODE> has the same meaning as the second
2234 argument to the Perl built-in C<flock> function; for example
2235 C<LOCK_SH> or C<LOCK_EX | LOCK_NB>. (These constants are provided by
2236 the C<use Fcntl ':flock'> declaration.)
2238 C<MODE> is optional; the default is C<LOCK_EX>.
2240 C<Tie::File> maintains an internal table of the byte offset of each
2241 record it has seen in the file.
2243 When you use C<flock> to lock the file, C<Tie::File> assumes that the
2244 read cache is no longer trustworthy, because another process might
2245 have modified the file since the last time it was read. Therefore, a
2246 successful call to C<flock> discards the contents of the read cache
2247 and the internal record offset table.
2249 C<Tie::File> promises that the following sequence of operations will
2252 my $o = tie @array, "Tie::File", $filename;
2255 In particular, C<Tie::File> will I<not> read or write the file during
2256 the C<tie> call. (Exception: Using C<mode =E<gt> O_TRUNC> will, of
2257 course, erase the file during the C<tie> call. If you want to do this
2258 safely, then open the file without C<O_TRUNC>, lock the file, and use
2261 The best way to unlock a file is to discard the object and untie the
2262 array. It is probably unsafe to unlock the file without also untying
2263 it, because if you do, changes may remain unwritten inside the object.
2264 That is why there is no shortcut for unlocking. If you really want to
2265 unlock the file prematurely, you know what to do; if you don't know
2266 what to do, then don't do it.
2268 All the usual warnings about file locking apply here. In particular,
2269 note that file locking in Perl is B<advisory>, which means that
2270 holding a lock will not prevent anyone else from reading, writing, or
2271 erasing the file; it only prevents them from getting another lock at
2272 the same time. Locks are analogous to green traffic lights: If you
2273 have a green light, that does not prevent the idiot coming the other
2274 way from plowing into you sideways; it merely guarantees to you that
2275 the idiot does not also have a green light at the same time.
2279 my $old_value = $o->autochomp(0); # disable autochomp option
2280 my $old_value = $o->autochomp(1); # enable autochomp option
2282 my $ac = $o->autochomp(); # recover current value
2284 See L<"autochomp">, above.
2286 =head2 C<defer>, C<flush>, C<discard>, and C<autodefer>
2288 See L<"Deferred Writing">, below.
2292 $off = $o->offset($n);
2294 This method returns the byte offset of the start of the C<$n>th record
2295 in the file. If there is no such record, it returns an undefined
2298 =head1 Tying to an already-opened filehandle
2300 If C<$fh> is a filehandle, such as is returned by C<IO::File> or one
2301 of the other C<IO> modules, you may use:
2303 tie @array, 'Tie::File', $fh, ...;
2305 Similarly if you opened that handle C<FH> with regular C<open> or
2306 C<sysopen>, you may use:
2308 tie @array, 'Tie::File', \*FH, ...;
2310 Handles that were opened write-only won't work. Handles that were
2311 opened read-only will work as long as you don't try to modify the
2312 array. Handles must be attached to seekable sources of data---that
2313 means no pipes or sockets. If C<Tie::File> can detect that you
2314 supplied a non-seekable handle, the C<tie> call will throw an
2315 exception. (On Unix systems, it can detect this.)
2317 Note that Tie::File will only close any filehandles that it opened
2318 internally. If you passed it a filehandle as above, you "own" the
2319 filehandle, and are responsible for closing it after you have untied
2322 =head1 Deferred Writing
2324 (This is an advanced feature. Skip this section on first reading.)
2326 Normally, modifying a C<Tie::File> array writes to the underlying file
2327 immediately. Every assignment like C<$a[3] = ...> rewrites as much of
2328 the file as is necessary; typically, everything from line 3 through
2329 the end will need to be rewritten. This is the simplest and most
2330 transparent behavior. Performance even for large files is reasonably
2333 However, under some circumstances, this behavior may be excessively
2334 slow. For example, suppose you have a million-record file, and you
2341 The first time through the loop, you will rewrite the entire file,
2342 from line 0 through the end. The second time through the loop, you
2343 will rewrite the entire file from line 1 through the end. The third
2344 time through the loop, you will rewrite the entire file from line 2 to
2347 If the performance in such cases is unacceptable, you may defer the
2348 actual writing, and then have it done all at once. The following loop
2349 will perform much better for large files:
2357 If C<Tie::File>'s memory limit is large enough, all the writing will
2358 done in memory. Then, when you call C<-E<gt>flush>, the entire file
2359 will be rewritten in a single pass.
2361 (Actually, the preceding discussion is something of a fib. You don't
2362 need to enable deferred writing to get good performance for this
2363 common case, because C<Tie::File> will do it for you automatically
2364 unless you specifically tell it not to. See L<"autodeferring">,
2367 Calling C<-E<gt>flush> returns the array to immediate-write mode. If
2368 you wish to discard the deferred writes, you may call C<-E<gt>discard>
2369 instead of C<-E<gt>flush>. Note that in some cases, some of the data
2370 will have been written already, and it will be too late for
2371 C<-E<gt>discard> to discard all the changes. Support for
2372 C<-E<gt>discard> may be withdrawn in a future version of C<Tie::File>.
2374 Deferred writes are cached in memory up to the limit specified by the
2375 C<dw_size> option (see above). If the deferred-write buffer is full
2376 and you try to write still more deferred data, the buffer will be
2377 flushed. All buffered data will be written immediately, the buffer
2378 will be emptied, and the now-empty space will be used for future
2381 If the deferred-write buffer isn't yet full, but the total size of the
2382 buffer and the read cache would exceed the C<memory> limit, the oldest
2383 records will be expired from the read cache until the total size is
2386 C<push>, C<pop>, C<shift>, C<unshift>, and C<splice> cannot be
2387 deferred. When you perform one of these operations, any deferred data
2388 is written to the file and the operation is performed immediately.
2389 This may change in a future version.
2391 If you resize the array with deferred writing enabled, the file will
2392 be resized immediately, but deferred records will not be written.
2393 This has a surprising consequence: C<@a = (...)> erases the file
2394 immediately, but the writing of the actual data is deferred. This
2395 might be a bug. If it is a bug, it will be fixed in a future version.
2397 =head2 Autodeferring
2399 C<Tie::File> tries to guess when deferred writing might be helpful,
2400 and to turn it on and off automatically.
2406 In this example, only the first two assignments will be done
2407 immediately; after this, all the changes to the file will be deferred
2408 up to the user-specified memory limit.
2410 You should usually be able to ignore this and just use the module
2411 without thinking about deferring. However, special applications may
2412 require fine control over which writes are deferred, or may require
2413 that all writes be immediate. To disable the autodeferment feature,
2416 (tied @o)->autodefer(0);
2420 tie @array, 'Tie::File', $file, autodefer => 0;
2423 Similarly, C<-E<gt>autodefer(1)> re-enables autodeferment, and
2424 C<-E<gt>autodefer()> recovers the current value of the autodefer setting.
2427 =head1 CONCURRENT ACCESS TO FILES
2429 Caching and deferred writing are inappropriate if you want the same
2430 file to be accessed simultaneously from more than one process. Other
2431 optimizations performed internally by this module are also
2432 incompatible with concurrent access. A future version of this module will
2433 support a C<concurrent =E<gt> 1> option that enables safe concurrent access.
2435 Previous versions of this documentation suggested using C<memory
2436 =E<gt> 0> for safe concurrent access. This was mistaken. Tie::File
2437 will not support safe concurrent access before version 0.98.
2441 (That's Latin for 'warnings'.)
2447 Reasonable effort was made to make this module efficient. Nevertheless,
2448 changing the size of a record in the middle of a large file will
2449 always be fairly slow, because everything after the new record must be
2454 The behavior of tied arrays is not precisely the same as for regular
2455 arrays. For example:
2457 # This DOES print "How unusual!"
2458 undef $a[10]; print "How unusual!\n" if defined $a[10];
2460 C<undef>-ing a C<Tie::File> array element just blanks out the
2461 corresponding record in the file. When you read it back again, you'll
2462 get the empty string, so the supposedly-C<undef>'ed value will be
2463 defined. Similarly, if you have C<autochomp> disabled, then
2465 # This DOES print "How unusual!" if 'autochomp' is disabled
2467 print "How unusual!\n" if $a[10];
2469 Because when C<autochomp> is disabled, C<$a[10]> will read back as
2470 C<"\n"> (or whatever the record separator string is.)
2472 There are other minor differences, particularly regarding C<exists>
2473 and C<delete>, but in general, the correspondence is extremely close.
2477 I have supposed that since this module is concerned with file I/O,
2478 almost all normal use of it will be heavily I/O bound. This means
2479 that the time to maintain complicated data structures inside the
2480 module will be dominated by the time to actually perform the I/O.
2481 When there was an opportunity to spend CPU time to avoid doing I/O, I
2482 usually tried to take it.
2486 You might be tempted to think that deferred writing is like
2487 transactions, with C<flush> as C<commit> and C<discard> as
2488 C<rollback>, but it isn't, so don't.
2492 There is a large memory overhead for each record offset and for each
2493 cache entry: about 310 bytes per cached data record, and about 21 bytes per offset table entry.
2495 The per-record overhead will limit the maximum number of records you
2496 can access per file. Note that I<accessing> the length of the array
2497 via C<$x = scalar @tied_file> accesses B<all> records and stores their
2498 offsets. The same for C<foreach (@tied_file)>, even if you exit the
2505 This version promises absolutely nothing about the internals, which
2506 may change without notice. A future version of the module will have a
2507 well-defined and stable subclassing API.
2509 =head1 WHAT ABOUT C<DB_File>?
2511 People sometimes point out that L<DB_File> will do something similar,
2512 and ask why C<Tie::File> module is necessary.
2514 There are a number of reasons that you might prefer C<Tie::File>.
2515 A list is available at C<http://perl.plover.com/TieFile/why-not-DB_File>.
2521 To contact the author, send email to: C<mjd-perl-tiefile+@plover.com>
2523 To receive an announcement whenever a new version of this module is
2524 released, send a blank email message to
2525 C<mjd-perl-tiefile-subscribe@plover.com>.
2527 The most recent version of this module, including documentation and
2528 any news of importance, will be available at
2530 http://perl.plover.com/TieFile/
2535 C<Tie::File> version 0.97 is copyright (C) 2003 Mark Jason Dominus.
2537 This library is free software; you may redistribute it and/or modify
2538 it under the same terms as Perl itself.
2540 These terms are your choice of any of (1) the Perl Artistic Licence,
2541 or (2) version 2 of the GNU General Public License as published by the
2542 Free Software Foundation, or (3) any later version of the GNU General
2545 This library is distributed in the hope that it will be useful,
2546 but WITHOUT ANY WARRANTY; without even the implied warranty of
2547 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
2548 GNU General Public License for more details.
2550 You should have received a copy of the GNU General Public License
2551 along with this library program; it should be in the file C<COPYING>.
2552 If not, write to the Free Software Foundation, Inc., 51 Franklin Street,
2553 Fifth Floor, Boston, MA 02110-1301, USA
2555 For licensing inquiries, contact the author at:
2559 Philadelphia, PA 19107
2563 C<Tie::File> version 0.97 comes with ABSOLUTELY NO WARRANTY.
2564 For details, see the license.
2568 Gigantic thanks to Jarkko Hietaniemi, for agreeing to put this in the
2569 core when I hadn't written it yet, and for generally being helpful,
2570 supportive, and competent. (Usually the rule is "choose any one.")
2571 Also big thanks to Abhijit Menon-Sen for all of the same things.
2573 Special thanks to Craig Berry and Peter Prymmer (for VMS portability
2574 help), Randy Kobes (for Win32 portability help), Clinton Pierce and
2575 Autrijus Tang (for heroic eleventh-hour Win32 testing above and beyond
2576 the call of duty), Michael G Schwern (for testing advice), and the
2577 rest of the CPAN testers (for testing generally).
2579 Special thanks to Tels for suggesting several speed and memory
2582 Additional thanks to:
2588 Jarkko Hietaniemi (again) /
2592 Tassilo von Parseval /
2598 Autrijus Tang (again) /
2604 More tests. (Stuff I didn't think of yet.)
2608 Fixed-length mode. Leave-blanks mode.
2610 Maybe an autolocking mode?
2612 For many common uses of the module, the read cache is a liability.
2613 For example, a program that inserts a single record, or that scans the
2614 file once, will have a cache hit rate of zero. This suggests a major
2615 optimization: The cache should be initially disabled. Here's a hybrid
2616 approach: Initially, the cache is disabled, but the cache code
2617 maintains statistics about how high the hit rate would be *if* it were
2618 enabled. When it sees the hit rate get high enough, it enables
2619 itself. The STAT comments in this code are the beginning of an
2620 implementation of this.
2622 Record locking with fcntl()? Then the module might support an undo
2623 log and get real transactions. What a tour de force that would be.
2625 Keeping track of the highest cached record. This would allow reads-in-a-row
2626 to skip the cache lookup faster (if reading from 1..N with empty cache at
2627 start, the last cached value will be always N-1).