(1) To select data. Filters use patterns (masks) against which all data are compared and only matching data are "passed through," hence the concept of a filter. For example, e-mail clients and servers can look for messages with text patterns that are recognized as spam and delete them. An e-mail client can be set up to filter messages and store them in separate mailboxes as a way of organizing the mail, or it can be set to alert the user when a certain type of message arrives. See Bayesian filtering and spam filter.

(2) To change data. In this usage of the term, the concept of a filter (a pass-through device) is less obvious, but nevertheless widely used. For example, a sort routine changes the sequence of data. A conversion routine (import or export filter) changes one type of data, text or graphics format into another. See image filter and parse.

(3) See packet filter.