aboutsummaryrefslogtreecommitdiffstats
path: root/db1-ast/hash/README
blob: f29ccf7e1ba608672c0472aa47fc999b4228c706 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
#	@(#)README	8.1 (Berkeley) 6/4/93

This package implements a superset of the hsearch and dbm/ndbm libraries.

Test Programs:
	All test programs which need key/data pairs expect them entered
	with key and data on separate lines

	tcreat3.c	
		Takes 
			bucketsize (bsize), 
			fill factor (ffactor), and
			initial number of elements (nelem).  
		Creates a hash table named hashtest containing the 
		keys/data pairs entered from standard in.
	thash4.c
		Takes
			bucketsize (bsize), 
			fill factor (ffactor), 
			initial number of elements (nelem)
			bytes of cache (ncached), and
			file from which to read data  (fname)
		Creates a table from the key/data pairs on standard in and
		then does a read of each key/data in fname
	tdel.c
		Takes
			bucketsize (bsize), and
			fill factor (ffactor).
			file from which to read data (fname)
		Reads each key/data pair from fname and deletes the
		key from the hash table hashtest
	tseq.c
		Reads the key/data pairs in the file hashtest and writes them
		to standard out.
	tread2.c
		Takes
			butes of cache (ncached).
		Reads key/data pairs from standard in and looks them up
		in the file hashtest.
	tverify.c
		Reads key/data pairs from standard in, looks them up
		in the file hashtest, and verifies that the data is
		correct.

NOTES:

The file search.h is provided for using the hsearch compatible interface
on BSD systems.  On System V derived systems, search.h should appear in 
/usr/include.

The man page ../man/db.3 explains the interface to the hashing system.
The file hash.ps is a postscript copy of a paper explaining
the history, implementation, and performance of the hash package.

"bugs" or idiosyncracies

If you have a lot of overflows, it is possible to run out of overflow
pages.  Currently, this will cause a message to be printed on stderr.
Eventually, this will be indicated by a return error code.

If you are using the ndbm interface and exit without flushing or closing the
file, you may lose updates since the package buffers all writes.  Also,
the db interface only creates a single database file.  To avoid overwriting
the user's original file, the suffix ".db" is appended to the file name
passed to dbm_open.  Additionally, if your code "knows" about the historic
.dir and .pag files, it will break.  

There is a fundamental difference between this package and the old hsearch.
Hsearch requires the user to maintain the keys and data in the application's
allocated memory while hash takes care of all storage management.  The down
side is that the byte strings passed in the ENTRY structure must be null
terminated (both the keys and the data).