Commit | Line | Data |
bf941eae |
1 | package DBM::Deep::Engine; |
2 | |
3 | use 5.006_000; |
4 | |
5 | use strict; |
6 | use warnings FATAL => 'all'; |
7 | |
8 | use DBM::Deep::Iterator (); |
9 | |
10 | # File-wide notes: |
11 | # * Every method in here assumes that the storage has been appropriately |
12 | # safeguarded. This can be anything from flock() to some sort of manual |
13 | # mutex. But, it's the caller's responsability to make sure that this has |
14 | # been done. |
15 | |
16 | # Setup file and tag signatures. These should never change. |
17 | sub SIG_FILE () { 'DPDB' } |
18 | sub SIG_HEADER () { 'h' } |
19 | sub SIG_HASH () { 'H' } |
20 | sub SIG_ARRAY () { 'A' } |
21 | sub SIG_NULL () { 'N' } |
22 | sub SIG_DATA () { 'D' } |
23 | sub SIG_INDEX () { 'I' } |
24 | sub SIG_BLIST () { 'B' } |
25 | sub SIG_FREE () { 'F' } |
26 | sub SIG_SIZE () { 1 } |
27 | |
64a531e5 |
28 | =head1 NAME |
29 | |
30 | DBM::Deep::Engine |
31 | |
32 | =head1 PURPOSE |
33 | |
34 | This is an internal-use-only object for L<DBM::Deep/>. It mediates the low-level |
35 | mapping between the L<DBM::Deep/> objects and the storage medium. |
36 | |
37 | The purpose of this documentation is to provide low-level documentation for |
38 | developers. It is B<not> intended to be used by the general public. This |
39 | documentation and what it documents can and will change without notice. |
40 | |
41 | =head1 OVERVIEW |
42 | |
43 | The engine exposes an API to the DBM::Deep objects (DBM::Deep, DBM::Deep::Array, |
44 | and DBM::Deep::Hash) for their use to access the actual stored values. This API |
45 | is the following: |
46 | |
47 | =over 4 |
48 | |
49 | =item * new |
50 | |
51 | =item * read_value |
52 | |
53 | =item * get_classname |
54 | |
55 | =item * make_reference |
56 | |
57 | =item * key_exists |
58 | |
59 | =item * delete_key |
60 | |
61 | =item * write_value |
62 | |
63 | =item * get_next_key |
64 | |
65 | =item * setup_fh |
66 | |
67 | =item * begin_work |
68 | |
69 | =item * commit |
70 | |
71 | =item * rollback |
72 | |
73 | =item * lock_exclusive |
74 | |
75 | =item * lock_shared |
76 | |
77 | =item * unlock |
78 | |
79 | =back |
80 | |
81 | They are explained in their own sections below. These methods, in turn, may |
82 | provide some bounds-checking, but primarily act to instantiate objects in the |
83 | Engine::Sector::* hierarchy and dispatch to them. |
84 | |
85 | =head1 TRANSACTIONS |
86 | |
87 | Transactions in DBM::Deep are implemented using a variant of MVCC. This attempts |
88 | to keep the amount of actual work done against the file low while stil providing |
89 | Atomicity, Consistency, and Isolation. Durability, unfortunately, cannot be done |
90 | with only one file. |
91 | |
92 | =head2 STALENESS |
93 | |
94 | If another process uses a transaction slot and writes stuff to it, then |
95 | terminates, the data that process wrote it still within the file. In order to |
96 | address this, there is also a transaction staleness counter associated within |
97 | every write. Each time a transaction is started, that process increments that |
98 | transaction's staleness counter. If, when it reads a value, the staleness |
99 | counters aren't identical, DBM::Deep will consider the value on disk to be stale |
100 | and discard it. |
101 | |
102 | =head2 DURABILITY |
103 | |
104 | The fourth leg of ACID is Durability, the guarantee that when a commit returns, |
105 | the data will be there the next time you read from it. This should be regardless |
106 | of any crashes or powerdowns in between the commit and subsequent read. |
107 | DBM::Deep does provide that guarantee; once the commit returns, all of the data |
108 | has been transferred from the transaction shadow to the HEAD. The issue arises |
109 | with partial commits - a commit that is interrupted in some fashion. In keeping |
110 | with DBM::Deep's "tradition" of very light error-checking and non-existent |
111 | error-handling, there is no way to recover from a partial commit. (This is |
112 | probably a failure in Consistency as well as Durability.) |
113 | |
114 | Other DBMSes use transaction logs (a separate file, generally) to achieve |
115 | Durability. As DBM::Deep is a single-file, we would have to do something |
116 | similar to what SQLite and BDB do in terms of committing using synchonized |
117 | writes. To do this, we would have to use a much higher RAM footprint and some |
118 | serious programming that make my head hurts just to think about it. |
119 | |
120 | =cut |
121 | |
122 | |
123 | |
bf941eae |
124 | =head2 get_next_key( $obj, $prev_key ) |
125 | |
126 | This takes an object that provides _base_offset() and an optional string |
127 | representing the prior key returned via a prior invocation of this method. |
128 | |
129 | This method delegates to C<< DBM::Deep::Iterator->get_next_key() >>. |
130 | |
131 | =cut |
132 | |
133 | # XXX Add staleness here |
134 | sub get_next_key { |
135 | my $self = shift; |
136 | my ($obj, $prev_key) = @_; |
137 | |
138 | # XXX Need to add logic about resetting the iterator if any key in the reference has changed |
139 | unless ( $prev_key ) { |
140 | $obj->{iterator} = DBM::Deep::Iterator->new({ |
141 | base_offset => $obj->_base_offset, |
142 | engine => $self, |
143 | }); |
144 | } |
145 | |
146 | return $obj->{iterator}->get_next_key( $obj ); |
147 | } |
148 | |
149 | 1; |
150 | __END__ |