btree
—
btree database access method
#include
<sys/types.h>
#include
<db.h>
The routine
dbopen
() is the library interface
to database files. One of the supported file formats is
btree
files. The general description of the
database access methods is in
dbopen(3),
this manual page describes only the
btree
specific information.
The
btree
data structure is a sorted,
balanced tree structure storing associated key/data pairs.
The
btree
access method specific data
structure provided to
dbopen
() is defined
in the
<db.h>
include file as follows:
typedef struct {
u_long flags;
u_int cachesize;
int maxkeypage;
int minkeypage;
u_int psize;
int (*compare)(const DBT *key1, const DBT *key2);
size_t (*prefix)(const DBT *key1, const DBT *key2);
int lorder;
} BTREEINFO;
The elements of this structure are as follows:
- flags
- The flag value is specified by or'ing any of
the following values:
R_DUP
- Permit duplicate keys in the tree, i.e., permit insertion if the key
to be inserted already exists in the tree. The default behavior, as
described in
dbopen(3),
is to overwrite a matching key when inserting a new key or to fail if
the
R_NOOVERWRITE
flag is
specified. The R_DUP
flag is
overridden by the R_NOOVERWRITE
flag, and if the R_NOOVERWRITE
flag
is specified, attempts to insert duplicate keys into the tree will
fail.
If the database contains duplicate keys, the order of retrieval of
key/data pairs is undefined if the
get routine is used, however,
seq routine calls with the
R_CURSOR
flag set will always
return the logical “first” of any group of duplicate
keys.
- cachesize
- A suggested maximum size (in bytes) of the memory cache. This value is
only advisory, and the access method will
allocate more memory rather than fail. Since every search examines the
root page of the tree, caching the most recently used pages substantially
improves access time. In addition, physical writes are delayed as long as
possible, so a moderate cache can reduce the number of I/O operations
significantly. Obviously, using a cache increases (but only increases) the
likelihood of corruption or lost data if the system crashes while a tree
is being modified. If cachesize is 0 (no
size is specified) a default cache is used.
- maxkeypage
- The maximum number of keys which will be stored on any single page. Not
currently implemented.
- minkeypage
- The minimum number of keys which will be stored on any single page. This
value is used to determine which keys will be stored on overflow pages,
i.e., if a key or data item is longer than the pagesize divided by the
minkeypage value, it will be stored on overflow pages instead of in the
page itself. If minkeypage is 0 (no
minimum number of keys is specified) a value of 2 is used.
- psize
- Page size is the size (in bytes) of the pages used for nodes in the tree.
The minimum page size is 512 bytes and the maximum page size is 64K. If
psize is 0 (no page size is specified) a
page size is chosen based on the underlying file system I/O block
size.
- compare
- Compare is the key comparison function. It must return an integer less
than, equal to, or greater than zero if the first key argument is
considered to be respectively less than, equal to, or greater than the
second key argument. The same comparison function must be used on a given
tree every time it is opened. If compare
is
NULL
(no comparison function is
specified), the keys are compared lexically, with shorter keys considered
less than longer keys.
- prefix
- The prefix element is the prefix
comparison function. If specified, this routine must return the number of
bytes of the second key argument which are necessary to determine that it
is greater than the first key argument. If the keys are equal, the key
length should be returned. Note, the usefulness of this routine is very
data dependent, but, in some data sets can produce significantly reduced
tree sizes and search times. If prefix is
NULL
(no prefix function is specified),
and no comparison function is specified, a
default lexical comparison routine is used. If
prefix is
NULL
and a comparison routine is
specified, no prefix comparison is done.
- lorder
- The byte order for integers in the stored database metadata. The number
should represent the order as an integer; for example, big endian order
would be the number 4,321. If lorder is 0
(no order is specified) the current host order is used.
If the file already exists (and the
O_TRUNC
flag is not specified), the values specified for the
flags,
lorder and
psize arguments are ignored in favor of the
values used when the tree was created.
Forward sequential scans of a tree are from the least key to the greatest.
Space freed up by deleting key/data pairs from the tree is never reclaimed,
although it is normally made available for reuse. This means that the
btree
storage structure is grow-only. The
only solutions are to avoid excessive deletions, or to create a fresh tree
periodically from a scan of an existing one.
Searches, insertions, and deletions in a
btree
will all complete in O lg base N
where base is the average fill factor. Often, inserting ordered data into
btree
s results in a low fill factor. This
implementation has been modified to make ordered insertion the best case,
resulting in a much better than normal page fill factor.
The
btree
access method routines may fail and
set
errno for any of the errors specified for
the library routine
dbopen(3).
dbopen(3),
hash(3),
mpool(3),
recno(3)
Douglas Comer,
The Ubiquitous B-tree, ACM Comput.
Surv. 11, 2, 121-138,
June 1979.
Bayer and
Unterauer, Prefix B-trees,
ACM Transactions on Database Systems,
1, Vol. 2,
11-26, March 1977.
D. E. Knuth,
The Art of Computer Programming Vol. 3: Sorting and
Searching, 471-480,
1968.
Only big and little endian byte order is supported.