12.29 dbm.fsdbm - File-system dbm

Module: dbm.fsdbm ¶

Implements fsdbm. Extends dbm.

Class: <fsdbm> ¶

{dbm.fsdbm} Fsdbm is a dbm implementation that directly uses the filesystem. Basically, it uses file names for keys, and file content for values. Unlike other dbm implementations, this doesn’t depend on external libraries—it is pure Scheme implementation—so it is always available, while other dbm implementations may not. Obviously, it is not suitable for the database that has lots of entries, or has entries deleted and added very frequently. The advantage is when the number of entries are relatively small, and the values are relatively large while keys are small. The database name given to <fsdbm> instance is used as a directory name that stores the data. The data files are stored in subdirectories under path of fsdbm instance, hashed by the key. Non-alphanumeric characters in the key is encoded like _3a for ’:’, for example. If a key is too long to be a file name, it is chopped to chunks, and each chunk but the last one is used as a directory name. Note that a long key name may still cause a problem, for example, some of old ’tar’ command can’t deal with pathnames (not each pathname components, but the entire pathname) longer than 256 characters.

Fsdbm implements all of the dbm protocol (see dbm - Generic DBM interface). It doesn’t have any fsdbm-specific procedures.