ed - Use the ed line-oriented text editor

You can prefix subcommands in ed with zero, one, or two addresses. These addresses let you refer to single lines or ranges of lines in the buffer. You do not need to specify addresses for certain subcommands that use default addresses. Consult the description for a particular subcommand. You can construct each address out of the following components:

.

The single dot character represents the current line in the buffer. Many subcommands set the current line; for example the e command sets it to the last line of the new file being edited.

$

The dollar sign refers to the last line in the buffer.

n

The number n refers to the nth line in the buffer.

/regexp/

This searches for a line containing a string that matches the regular expression, regexp. The search begins at the line immediately following the current line. It proceeds forward through the buffer; if ed reaches the end of the buffer without finding a match, it wraps around to the first line of the buffer and continues the search. If ed does not find a match, the search ends when it reaches the original current line. If it does find a match, the address /regexp/ refers to the first matching line. If you omit regexp, the last used regular expression becomes the object of the search. You can omit the trailing /. Within regexp, \/ represents a literal slash and not the regexp delimiter.

?regexp?

This is similar to the previous address form, except that the search goes backward through the buffer. If the search reaches the first line in the buffer without finding a match, ed wraps around and continues searching backward from the last line in the buffer. If you omit regexp, the last used regular expression becomes the object of the search. You can omit the trailing ?. Within regexp, \? represents a literal question mark and not the regexp delimiter.

'l

The address is the line that is marked with the mark name l. The name l must be a lowercase letter that is set by the k subcommand.

You can specify address ranges in the following ways:

a1,a2

Specifies a range of addresses from address a1 to address a2, inclusive. If you omit a1 and a2 (that is, the comma alone is specified), this is equivalent to the range 1,$.

a1;a2

Is similar to the previous form except that ed resets the current line after calculating the first address, a1, so that the second address, a2, is relative to a1. If you omit a1 and a2 (that is, the semicolon alone is specified), this is equivalent to .;$. If you specify only a1 and the command requires both a1 and a2, the command operates as though you specified a range of:

a1;. command

>

Is equivalent to .,.+22 (that is, page forward), except that it never attempts to address any line beyond $.

<

Is equivalent to .-22,. (that is, page backward), except that it never addresses any line before line 1.

Subcommands generally take a maximum of zero, one, or two addresses, depending upon the particular subcommand. In the following descriptions, we show commands with their default addresses (that is the addresses used when you don't specify any addresses) in a form that shows the maximum number of permitted addresses for the command. In any of the subcommands that take a file argument, file can be a path name or:

!command-line

If you use the ! form, ed runs the given command line, reading its standard output (stdout) or writing its standard input (stdin), depending on whether the ed command does reading or writing.

If a terminal disconnect is detected:

• If the buffer is not empty and has changed since the last write, the ed utility will attempt to write a copy of the buffer to a file named ed.hup in the current directory. If this write fails, ed will attempt to write a copy of the buffer to a file name ed.hup in the directory that is named by the HOME environment variable. If both these attempts fail, ed will exit without saving the buffer.
• The ed utility will not write the file to the currently remembered path name or return to command mode, and will terminate with the exit status of 1.

If an end-of-file is detected on standard input:

• If the ed utility is in input mode, ed will terminate input mode and return to command mode. Any partially entered lines (that is, input text without a terminating newline) will be saved.
• If the ed utility is in command mode, it will act as if a q command had been entered.

ed accepts the following subcommands:

.a

Appends text after the specified line. Valid addresses range from 0 (text is placed at the beginning of the buffer, before the first line) to $ (text is placed after the last line of the buffer). ed reads lines of text from the workstation until a line consisting solely of an unescaped . (dot) is entered. ed sets the current-line indicator to the last line appended.

.,.c

Changes the addressed range of lines by deleting the lines and then reading new text in the manner of the a or i subcommands. If the variable _UNIX03 is set to YES, address 0 is valid for this subcommand and it will be interpreted as if address 1 were specified.

.,.d

Deletes the addressed range of lines. The line after the last line deleted becomes the new current line. If you delete the last line of the buffer, ed sets the current line to the new last line. If no lines remain in the buffer, it sets the current line to 0.

E[file]

Is similar to the e command, but ed gives no warning if you have changed the buffer.

e [file]

Replaces the contents of the current buffer with the contents of file. The text conversion that is specified on the ed command (for example, the -B or -W option) is used. If you did not specify file, ed uses the remembered file name, if any. In all cases, the e subcommand sets the remembered file name to the file that it has just read into the buffer. ed displays a count of the bytes in the file unless it is in quiet mode. If you have changed the current buffer since the last time its contents were written, ed warns you if you try to run an e subcommand, and does not run the subcommand. If you enter the e subcommand a second time, ed goes ahead and runs the command.

f [file]

Changes the remembered file name to file. ed displays the new remembered file name. If you do not specify file, ed displays the current remembered file name.

1,$G/regexp/

Is similar to the g command except that when ed finds a line that matches regexp, it prints the line and waits for you to type in the subcommand to be run. You cannot use the a, c, i, g, G, v, and V subcommands. If you enter &, the G subcommand reruns the last subcommand you typed in. If you just press <Enter>, G does not run any subcommand for that line. Note that the subcommands input as part of the execution of the G subcommand can address and affect any lines in the buffer. If the variable _UNIX03 is set to YES, any line modified by the subcommand will be unmarked.

1,$g/regexp/command

Performs command on all lines that contain strings matching the regular expression regexp. This subcommand works in two passes. In the first pass, ed searches the given range of lines and marks all those that contain strings matching the regular expression regexp. The second pass performs command on those lines. If the variable _UNIX03 is set to YES, any line modified by the command will be unmarked. You cannot use !, g, G, V, or v as command. command consists of one or more ed subcommands, the first of which must appear on the same line as the g subcommand. All lines of a multiline command list, except the last, must end with a backslash (\). If command is empty, ed assumes it to be the p subcommand. If no lines match regexp, ed does not change the current line number; otherwise, the current line number is the one set by the last subcommand in command. Instead of the slash (/) to delimit regexp, you can use any character other than space or newline.

H

Tells ed to display more descriptive messages when errors occur. If ed is already printing descriptive messages, H returns to terse error messages. Normally, ed indicates error messages by displaying a ?. When you turn on descriptive error messages with this subcommand, ed also displays the descriptive message for the most recent ? message.

h

Provides a brief explanation of the last error that occurred. This does not change the current line number.

.i

Works similarly to the a subcommand, except that ed places the text before the addressed line. Valid addresses range from line 1 to $ (the last line). ed sets the current line number to the last inserted line. If the variable _UNIX03 is set to YES, address 0 is valid for this subcommand and it will be interpreted as if address 1 were specified.

.,.+1j

Joins a range of lines into one line. To be precise, the j command removes all newline characters from the addressed range of lines, except for the last one. ed sets the current line number to the resulting combined line.

.kx

Marks the addressed line with the mark name x, which is any single lowercase letter of the alphabet. This lets you refer to a marked line with the construct 'x. This is called an absolute address, because it always refers to the same line, regardless of changes to the buffer.

.,.l

Displays the addressed range of lines, representing nonprintable (control) characters in a visible manner. The end of each line will be marked with a '$' character. The characters listed in the Base Definitions volume of IEEE Std 1003.1-2001, Table 5-1, Escape Sequences and Associated Actions ( '\\', '\a', '\b', '\f', '\r', '\t', '\v' ) shall be written as the corresponding escape sequence; the '\n' in that table is not applicable. If the variable _UNIX03 is set to YES,'$' characters within the text will be written with a preceding backslash. ed sets the current line to the last line so displayed. You can append this subcommand to most other commands, to check on the effect of those subcommands.

.,.ma

Moves the addressed lines to the point immediately following the line given by the address a. The address a must not be in the range of addressed lines. If address a is 0, ed moves the lines to the beginning of the buffer. The last line moved becomes the new current line.

.,.n

Displays the addressed lines in a way similar to the p command, but ed puts the line number and a tab character at the beginning of each line. The last line displayed becomes the new current line. You can append n to any subcommand (except for E, e, f, Q, r, w, or !) so that you can check on the effect that the subcommands had.

P

Turns on subcommand prompting if it is not already on. If you specified the -p prompt option on the ed command line, ed displays the prompt string whenever it is ready for you to type in another subcommand. If you did not include the -p option, ed uses the * character as a prompt. If subcommand prompting is currently turned on, issuing the P subcommand turns it off.

.,.p

Displays (prints) the addressed lines. The last line displayed becomes the new current line. You can append p to most subcommands, so that you can check on the effect that the subcommands had.

You can append p to any subcommand (except for E, e, f, Q, r, w, or !) so that you can check on the effect that the subcommands had.

Q

Quits unconditionally, without checking for buffer changes.

q

Causes the editor to exit. If you have made changes to the buffer since the last save and you try to quit, ed issues a warning. Entering the q subcommand again lets you quit, regardless of unsaved changes.

$r [file]

Reads the contents of the file into the buffer after the addressed line. The text conversion that is specified on the ed command (for example, the -B or -W option) is used. If the address is 0, ed places the text before the first line in the buffer. If you do not specify file, ed uses the remembered file name; if no remembered file name exists, file becomes the new remembered name. If file contains bytes that are not valid in the current character set, they are replaced by the rubout character.

The r subcommand displays the number of bytes read from file unless you specified the -s option. The last line read from the file becomes the new current line. If file is replaced by !, the rest of the line is considered a shell command line, the output of which is to be read.

.,.s/regexp/new/[flags]

Searches the specified range of lines for strings matching the regular expression regexp. Normally the s subcommand replaces the first such matching string in each line with the string new. The s subcommand sets the current line to the last line on which a substitution occurred. If ed makes no such replacements, ed considers it an error.

flags can be one of the following:

n

Replaces the nth matching string in the line instead of the first one.

g

Replaces every matching string in each line, not just the first one.

l

Displays the new current line in the format of the l subcommand.

n

Displays the new current line in the format of the n subcommand.

p

Displays the new current line in the format of the p subcommand.

You can use any single printable character other than space or newline instead of / to separate parts of the subcommand provided that you use the same character to delimit all parts of the subcommand. You can omit the trailing delimiter.

You can include a newline in the new string by putting a \ immediately in front of the newline. This is a good way to split a line into two lines. If new consists only of the % character, s uses the new string from the previous s command. If the variable _UNIX03=YES is set and there was no previous s command, the use of % in this manner is an error. If & appears anywhere in new, ed replaces it with the text matching the regexp. If you want new to contain a literal ampersand, or percent sign, put a backslash (\) in front of the & or % character.

.,.ta

Copies the addressed lines to the point after the line given by the address a. The address a must not fall in the range of addressed lines. If address a is 0, ed copies the lines to the beginning of the buffer. This sets the current line to the last line copied.

u

Rolls back the effect of the last subcommand that changed the buffer. For the purposes of u, subcommands that change the buffer are: a, c, d, g, G, i, j, m, r, s, t, v, V, and (of course) u. This means that typing u repeatedly switches the most recent change back and forth. This subcommand sets the current line number to the value it had immediately before the subcommand being undone started.

1,$V/regexp/

Is similar to the G subcommand, except that this subcommand gives you the chance to edit only those lines that do not match the given regular expression.

1,$v/regexp/commands

Is similar to the g (global) command, except that ed applies the given commands only to lines that do not match the given regular expression.

1,$W [file]

Is similar to the w subcommand, except that this command appends data to the given file if the file already exists.

1,$w [file]

Writes the addressed lines of the buffer to the named file. The text conversion that is specified on the ed command (for example, the -B or -W option) is used. This does not change the current line. If you do not provide file, ed uses the remembered file name; if there is no remembered file name, file becomes the remembered name. If the output file does not exist, ed creates it. ed displays the number of characters written unless you had specified the -s option.

X

Prompts you to enter an encryption key. All subsequent e, r, and w subcommands use this key to decrypt or encrypt text read from or written to files. To turn encryption off, issue an X subcommand and press <Return> in response to the prompt for an encryption key.

!command

Runs command as if you typed it to your chosen command interpreter. If command contains the % character, ed replaces it with the current remembered file name. If you want a subcommand to contain a literal %, put a backslash (\) in front of the character. As a special case, typing !! reruns the previous command.

$=

Displays the line number of the addressed line. This does not change the current line.

.=

Displays the current line number.

.+1.,.+1

If you supply zero, one, or two addresses without an explicit subcommand, ed displays the addressed lines in the mode of the last print subcommand: p, l, or n. This sets the current line number to the last line displayed.

ed uses the following environment variables:

COLUMNS

Contains the terminal width in columns. ed folds lines at that point. If it is not set, ed uses the appropriate value from the terminfo database or if that is not available, it uses a default of 80.

HOME

Contains the path name of your home directory.

SHELL

Contains the full path name of the current shell.

TMPDIR

The path name of the directory being used for temporary files. If it is not set, ed uses /tmp.

_TEXT_CONV

Contains text conversion information for the command. The text conversion information is not used when either the -B option or the filecodeset or pgmcodeset option (-W option) is specified. For more information about text conversion, see Controlling text conversion for z/OS UNIX shell commands.

_UNIX03

For more information about the effect of _UNIX03 on this command, see Shell commands changed for UNIX03.

ed uses the following files:

/tmp/e*

This is the paging file. It holds a copy of the file being edited. You can change the directory for temporary files using the environment variable TMPDIR.

ed.hup

When ed receives a hang up signal (or detects a terminal disconnect) and the current buffer has changed since the last write, ed will attempt to write the current buffer to ed.hup in the current directory. If this write fails, ed will attempt to write the current buffer to ed.hup in the $HOME directory.

ed uses the following localization environment variables:

• LANG
• LC_ALL
• LC_COLLATE
• LC_CTYPE
• LC_MESSAGES
• LC_SYNTAX
• NLSPATH

Destination cannot straddle source in m and t

The range of lines being moved or copied by m or t cannot include the destination address.

Global command too long

There is a limit on the length of a global instruction (g or v). See Limits for this limit.

m and t require destination address

You must follow the m or t subcommands with an address indicating where you want to move or copy text. You omitted this address.

No remembered file name

You tried to run a subcommand that used a remembered file name (for example, you used w to write without specifying an output file name). However, there is no remembered file name at present. Run the subcommand again, but specify a file name this time.

Restricted shell

The command line invoked the restricted form of ed, but you tried an action that was not allowed in the restricted editor (the ! subcommand).

Temporary file error

You ran out of space on disk or encountered other errors involving the page file stored in the temporary file.

Warning: file not saved

You entered a subcommand to quit editing the current file, for example, q or e to edit a new file; however, you have changed the file since the last time you saved it. ed is suggesting that you save the file before you exit it; otherwise, your recent changes will be lost. To save the file, use the w command. If you really do not want to save the recent changes, use q to quit or e to edit a new file.

?file

An error occurred during an attempt to open or create file. This is applicable to the e, r, and w subcommands.

?

An unspecified error occurred. Use the h or H subcommand for more information. If the input to ed comes from a script rather than from a workstation, ed exits when any error occurs.