bfs -- scan big files


bfs [ - ] name


bfs is like ed(C) except that it is read-only and processes much larger files. Files can be up to 1024KB and 32K lines, with up to 255 characters per line. bfs is usually more efficient than ed for scanning a file, since the file is not copied to a buffer. It is most useful for identifying sections of a large file where csplit(C) can be used to divide it into more manageable pieces for editing.

Normally, the size of the file being scanned is printed, in the same way as the size of any file written with the w command. The optional dash (-) suppresses printing of sizes. Input is prompted for with an asterisk (*) when ``P'' and <Return> are typed. The ``P'' acts as a toggle, so prompting can be turned off again by entering another ``P'' and <Return>. Note that messages are given in response to errors only if prompting is turned on.

All address expressions described under ed are supported. In addition, regular expressions may be surrounded with two symbols other than the standard slash (/) and ``?'': A greater-than sign (>) indicates downward search without wraparound, and a less-than sign (<) indicates upward search without wraparound. Note that parentheses and curly braces are special and need to be escaped with a backslash (\). Since bfs uses a different regular expression-matching routine from ed, the regular expressions accepted are slightly wider in scope (see regex(S)). Differences between ed and bfs are listed below:

A regular expression followed by ``+'' means ``one or more times''. For example, [0-9]+ is equivalent to [0-9][0-9]*.

\{m\} \{m,\} \{m,u\}
Integer values enclosed in \{ \} indicate the number of times the preceding regular expression is to be applied. m is the minimum number and u is a number, less than 256, which is the maximum. If only m is present (for example, \{m\}), it indicates the exact number of times the regular expression is to be applied. \{m,\} is analogous to \{m,infinity\}. The plus (+) and star (*) operations are equivalent to \{1,\} and \{0,\} respectively.

The value of the enclosed regular expression is to be returned. The value will be stored in the (n+1)th argument following the subject argument. At most ten enclosed regular expressions are allowed. regex makes its assignments unconditionally.

Parentheses are used for grouping. An operator, for example *, +, \{ and \}, can work on a single character or a regular expression enclosed in parentheses. For example,

\ (a*\ (cb+\ )*\ )$0.
There is also a slight difference in mark names: only the letters ``a'' through ``z'' may be used, and all 26 marks are remembered.

The e, g, v, k, p, q, w, =, ! and null commands operate as described under ed except that e does not remember filenames and g and v, when given no arguments, return the line after the line you were on. Commands such as ---, +++-, +++=, -12, and +4p are accepted. Note that 1,10p and 1,10 will both print the first ten lines. The f command only prints the name of the file being scanned; there is no remembered filename. The w command is independent of output diversion, truncation, or crunching (see the xo, xt and xc commands, below). The following additional commands are available:

xf file
Further commands are taken from the named file. When an end-of-file is reached or an interrupt signal is received, or an error occurs, reading resumes with the file containing the xf. xf commands may be nested to a depth of 10.

xo [ file ]
Further output from the p and null commands is diverted to the named file. If file is missing, output is diverted to the standard output. Note that each diversion causes truncation or creation of the file.

: label
This positions a label in a command file. The label is terminated by a newline, and blanks between the ``:'' and the start of the label are ignored. This command may also be used to insert comments into a command file, since labels need not be referenced.

( . , . )xb/regular expression/label
A jump (either upward or downward) is made to label if the command succeeds. It fails under any of the following conditions:
On success, dot (.) is set to the line matched and a jump is made to label. This command is the only one that does not issue an error message on bad addresses, so it may be used to test whether addresses are bad before other commands are executed. Note that the command xb/^/ label is an unconditional jump.

The xb command is allowed only if it is read from somewhere other than a terminal. If it is read from a pipe only a downward jump is possible.

xt number
Output from the p and null commands is truncated to a maximum of number characters. The initial number is 255.

xv[ digit ] [ spaces ] [ value ]
The variable name is the specified digit following the xv. xv5100 or xv5 100 both assign the value 100 to the variable 5. xv61,100p assigns the value 1,100p to the variable 6. To reference a variable, put a ``%'' in front of the variable name. For example, using the above assignments for variables 5 and 6:


prints the first 100 lines.


globally searches for the characters ``100'' and prints each line containing a match. To escape the special meaning of ``%'', a ``&'' must precede it. For example,


could be used to match and list lines containing printf characters, decimal integers, or strings.

Another feature of the xv command is that the first line of output from a UNIX command can be stored into a variable. The only requirement is that the first character of value be a ``!''. For example,

xv5!cat junk
!rm junk
!echo "%5"
xv6!expr %6 + 1

puts the current line in variable 5, prints it, and increments the variable 6 by 1. To escape the special meaning of ``!'' as the first character of value, precede it with a ``\''. For example,


stores the value !date into variable 7.

xbz label

xbn label
These two commands test the last saved return code from the execution of a UNIX command !command) or nonzero value, respectively, and jump to the specified label. The two examples below search for the next five lines containing the string size:

: l
xv5!expr %5 - 1
!if 0%5 != 0 exit 2
xbn l
: l
xv4!expr %4 - 1
!if 0%4 = 0 exit 2
xbz l

xc [ switch ]
If switch is 1, output from the p and null commands is crunched; if switch is 0, it is not. Without an argument, xc reverses switch. Initially switch is set for no crunching. Crunched output has strings of tabs and blanks reduced to one blank and blank lines suppressed.


``?'' for errors in commands if prompting is turned off. Self-explanatory error messages when prompting is on.

See also

csplit(C), ed(C), umask(C)
© 2003 Caldera International, Inc. All rights reserved.
SCO OpenServer Release 5.0.7 -- 11 February 2003