Section: User Commands (1)
Updated: 2015
Index Return to Main Contents


smenu - filter that allows one to interactively select a word from stdin and outputs the selection to stdout.  


smenu [-h|-?] [-f configuration_file] \
      [-n lines] [-t [cols]] [-k] \
      [-s pattern] [-m message] [-w] \
      [-d] [-M] [-c] [-l] [-r] [-b] \
      [-a (i|e|c|b|s|t|ct|sf|st|mf|mt|sfe|ste|mfe|mte|da):ATTR]... \
      [-i regex] [-e regex] \
      [-C [i|e]<col selectors>] \
      [-R [i|e]<row selectors>] \
      [-S /regex/string/[g][v][s][i]] \
      [-I /regex/string/[g][v][s][i]] \
      [-E /regex/string/[g][v][s][i]] \
      [-A regex] [-Z regex] \
      [-N [regex]] [-U [regex]] [-F] [-D sub-option...] \
      [-1 regex [ATTR]] [-2 regex [ATTR]] ... [-5 regex [ATTR]] \
      [-g [string]] [-q] [-W bytes] [-L bytes] \
      [-T [separator]] [-P [separator]] [-p] \
      [-V] [-x|-X type [word] delay] [-/ prefix|substring|fuzzy] \

      <col selectors> ::= col1[-col2],...|<RE>,...
      <row selectors> ::= col1[-col2],...|<RE>,...
      <sub-option>    ::= [l|r:<char>]|[a:left|right]|[p:included|all|
      <ATTR>          ::= [fg][/bg][,style]
      <RE>            ::= <char>regex<char>

      <col/row selectors> and <RE> can be freely mixed.
      The parameters of -a and -D must be delimited by blanks.


This small utility acts as a filter when no input file is given (reads from stdin and writes to stdout) or takes its inputs from that file.

All read words are presented in a scrolling window on the terminal at the current cursor position without clearing the screen before.

The selection cursor is initially positioned on the first selectable word by default.

Options exists to explicitly or implicitly include or exclude some words by using extended regular expressions.

Notice that when some words are explicitly excluded they can no more be re-included after.

Excluded words are skipped when the selection cursor is moved and cannot be searched for.

The -W option can be used to set the characters (or multibyte sequences) which will be used to delimit the input words. The default delimiters are: SPACE, \t and \n.

The -L has a similar meaning for lines.

Special character sequences formed by a \ followed by one of the characters a b t n v f r and \ are understood and have their traditional meanings.

UTF-8 sequences introduced by \u are also understood. \u can be followed by 2,4,6 or 8 hexadecimal characters. An invalid UTF-8 sequence will be replaced by a dot (.), see also below.

Example: \uc3a9 means latin small letter e with acute.

Note that with most shells, the \ before the u need to be protected or escaped.

Quotations (single and double) in the input stream can be used to ignore the word separators so that a group of words are taken as a single entity.

Non printable characters in words that are not delimiters are converted to their traditional form (\n for end-of-line, \t for tabulation...) by default. A single dot (.) is also used as a placeholder otherwise.

Words containing only spaces, entered directly or resulting from a substitution, are also rejected unless they are not selectable. This allows special effects like creating blank lines for example. These words are also kept in column mode, selectable or not.

Warning, UTF-8 encoded codepoints are quietly converted into dots (.) when the user locale is not UTF-8 aware like POSIX or C by example.


Moving among words

The cursor can be moved in every direction by using the keyboard arrow keys (,,,) or the vi direction keys (h, j, k and l). HOME, END, PgDn and PgUp can also be used, if available, and have the following meanings:
←, hPrevious word
↑, kPrevious line
PgUp, KPrevious pages
HomeFirst word of the window
CTRL+Home, SHIFT+Home, CTRL+kFirst word
→, lNext Word
↓, jNext line
PgDn, JNext pages
EndLast word of the window
CTRL+End, SHIFT+End, CTRL+jLast word

If the -N, -U or -F are used then it is possible to directly access a word by entering its number. The numbering created using these option is done before any words substitution done using -S, -I or -E.

Using a combination of these options, it is easy to control which words will be numbered by adding a special symbol in it before using smenu and removing it (substituted by nothing) afterward using -I by example.

-E gives another way to do that, see below or more.  

Searching for a word

A word can be searched using different algorithms: prefix, substring of fuzzy.
prefix (keys ^ or =):
The sequence of characters entered must match the beginning of a word.
substring (keys " or '):
The sequence of characters entered must match a substring in a word.
fuzzy (keys ~ or *):
All the character in the entered sequence must appear in the same order in a word must not necessarily be consecutive.

The case is also ignored.

The cursor is placed, if possible, on the first matching word having the minimum number of gaps between the first and last matching character, see the difference between the actions of the s/S and n/N keys below.

This method also tolerates intermediate symbols not appearing in the words which will be ignored. If this is the case, the attributes of the approximatively matching words are changed into an error versions of them to warn the user to this situation.

The erroneous symbols will not be inserted in the search buffer.

By example: if the word abcdef is present in the standard input, then entering abxcdye puts abcdef in the search buffer and the word is added to the list of matching words and displayed with an error attribute (in red by default).

This special state will persist until all the symbols following the first erroneous one are deleted (using backspace) or if ESC is pressed.

During a search session, the cursor changes and each character entered is added in (or removed from) the search buffer. The display is refreshed after each change in this buffer.

The / key can also be used instead of any of these keys. By default is is programmed to do a fuzzy search but this can be altered by using the command line option (-/) or by tuning a configuration file, see below.

All the words matching the current search buffer are enhanced: The characters present in the current search buffer are highlighted in one way and the other characters in another way. Both of these highlighting methods are configurable.

Typically, if the user has entered the search sequence: o, s, then the matching word "words" will be displayed as words when the fuzzy algorithm is in use depending of the display attributes configured.

ESC can be used anytime to abort the current search session. ENTER and all cursor moves also terminate the search session but do not clear the list of the matchng words.

The user can then use the n/s/SPACE keys (forward) and the N/S keys (backward) to navigate in the list of matching words,

In fuzzy search mode, the s/S keys attempt to move the cursor to the next/previous word whose matching part forms a substring of this word. If no such matches exist, s/S and n/N do the same things. To move the cursor to the next/previous fuzzy match, use the n/N/SPACE keys. s means next substring match in this context while n just means next match.

If the user hits the Home or End key during a search session then the list of matching words is reduced to the words starting (respectively) ending with the current search pattern and the window is refreshed. For those who consider Home and End as non-intuitive, the CTRL-A and Ctrl-Z keys are also available in search mode as an alternative.

This behaviour is persistent until the user hit the ESC or ENTER key.

By example, if the search pattern in substring mode is sh and the user hits End, then only the words ending with sh will be added in the searched word list and enhanced.

Note that when a matching word is selected, its enhanced characters only show one of the multiple matching possibilities.

When not in a search session ESC can be also used to clear the list of matching words and to reset the search buffer.

In summary, here is the meaning of the special keys in search mode:
Keys which clear the list of matching words.

EscCancel searchYes

Keys which keep or update the list of matching words.

Previous wordYes
Previous lineYes
PgUpPrevious pageYes
CTRL+Home, SHIFT+Home, CTRL+kFirst wordYes

Next wordYes
Next lineYes
PgDnNext pagesYes
CTRL+End, SHIFT+End, CTRL+jLast wordYes

Home,CTRL-A Only keep the words starting with the search pattern No
End,CTRL-Z Only keep the words ending with the search pattern No

InsTag wordNo
DelUntag wordNo

Note that the search buffer is persistent as long as the same search algorithm is used and ESC has not been pressed.  

Selection and Exit

Pressing q gives the possibility to exit without selecting anything.

By default, ENTER writes the selected word to stdout when not in search mode otherwise it exits from this mode and does nothing more. If you want to be able to select a word even when in search mode, use the -r option to change this behavior.  

Tagging (multi-selections)

When the tagging is activated by using the command line -T or -P option, then the keys t, T, INS and DEL can be used to tag/untag some words. These tagged words will then be output on the standard output when ENTER is pressed.
Tag/untag or Pin/unpin the word under the cursor (toggle).
Tag or pin the matching words if any.
Untag or unpin the matching words if any.
Tag or pin the word under the cursor.
Untag or unpin the word under the cursor.


A small help message can be displayed when hitting ?. This display will last for 10s or until a valid key or ESC is pressed.  

Scroll bar

A scroll bar is displayed at the right of the scrolling window. Its appearance is meant to be classical but it has some particularities:
The scroll bar is not displayed if all the input words fit on only one line.
Otherwise, the scroll bar is always displayed except when the -q option is set. This option completely disables the scroll bar display.
When the scrolling window has only one line, the scroll bar has only 3 states:
v when on all but the last line, indicating that you can go down to see more.
^ when on the last line.
| otherwise.
When there is more than one line to display, / means that the window displays the first line, \ the last line. | is used to fill the gap, see below the different possible configurations.
\@\@^^\ Do not remove this trailing space!

A + can also appear in the scroll bar in lieu of the vertical bar, giving the relative position of the cursor line in the bunch of input words.  

Terminal resizing (also see BUGS/LIMITATIONS)

The windows is redrawn if the terminal is resized. The redrawing is actually done only 1s after the end of the resizing to avoid artefacts on screen. The cursor will remain on the current selected word but may be displayed at another place in the window.  

Unicode support

This utility is Unicode aware and should be able to display correctly any Unicode character (even double-width ones) as long as the current encoding is UTF-8 (UTF-8 in the output of the locale command).  


If a file with adequate permissions and the same name as the executable but prefixed with a dot is present in the current directory or in the user's home directory, then it will be parsed as a ini file. The values read from the file in the home directory will be overridden by the ones read from the local directory (if it is present).

Missing and bad keywords are silently skipped.

The values read, if valid, override the default hard-coded ones.

If a value is invalid an error message is shown and the program terminates.

The values of the timers must be given in units of 1/10 of a second.

Here is an example giving the syntax and the names of the keywords allowed:

  ; The terminal must have at least 8 colors and/or have attributes like bold
  ; and reverse for this to be useful
  ; if not the following settings will be ignored.

  method=ansi             ; classic | ansi (default)

  cursor=0/2              ; cursor attributes
  cursor_on_tag=0/2,u     ; cursor on tag attributes
  shift=6,b               ; shift symbol attributes
  bar = 7/4,b             ; scroll bar attributes
  search_field = 0/6      ; search field attributes
  search_text = 7,bu      ; search text attributes
  match_field = 1,b       ; matching words field attributes
  match_text = 7,bu       ; matching words text attributes
  search_err_field = 1    ; approximate search field attributes
  search_err_text = 1,r   ; approximate search text attributes
  ; match_err_field = 3   ; approximate matching words field attributes
  match_err_text = 1      ; approximate matching words text attributes
  ; include = b           ; selectable color attributes
  exclude = 4/0,u         ; non-selectable color attributes
  tag = 0/5               ; tagged (selected) attributes
  daccess = 3,b           ; direct access tag attributes

  special1 = 7/4,b        ; attributes for the special level 1
  special2 = bu           ; attributes for the special level 2
  special3 = /3,b         ; attributes for the special level 3
  special4 = 7/4          ; attributes for the special level 4
  special5 = 7/2,b        ; attributes for the special level 5

  lines = 7               ; default number of lines of the window

  word_length = 1024      ; arbitrary max length of input words (int)
  words = 32767           ; arbitrary max number of allowed input
                          ; words (int)
  columns = 128           ; arbitrary max number of columns (int)

  search = 60             ; search timers in 1/10 s
  help = 150              ; duration of the help message in 1/10 s
  window = 7              ; delay before redrawing if the size of the
                          ; terminal's window change in 1/10 s
  direct_access = 6       ; duration allowed to add a new digit to
                          ; the direct word access number in 1/10 s

  default_search_method = substring

The method keyword can take the two possible values displayed above and determines if you want to use the native method (limited to 8 colors) of the ansi method (ISO 8613-6) if your terminal supports more than 8 colors.

The default value corresponds to ansi.

The attributes syntax is [fg][/bg][,toggles] where fg and bg are numbers representing the foreground and background color and toggles is a strings which can contain the characters b, d, r, s, u and i standing for bold, dim, reverse, standout, underline and italic.

Spaces are allowed anywhere in the lines and between them, even around the =.
Everything following a ; is ignored.
When undefined, the default limits are:


-h or -?
Displays a long (-h) or short (-?) help message and exits.
-f configuration_file
This option gives the possibility to select an alternative configuration file. If the given file doesn't exist or is not readable then the default values will be used.

The .smenu files in the user's home directory and in the current directory, if present, will be ignored when this option is used.

-n lines
Gives the maximum number of lines in the scrolling selection window. By default five lines at most are displayed and the other ones, if any, need you to scroll the window.

The special value 0 sets this number to match the number of lines in the terminal (minus the lines taken by the message if any). This remains true even if the terminal is resized.

-t [columns]
This option sets the tabulation mode and, if a number is specified, attents to set the number of displayed columns to that number. In this mode, embedded line separators are ignored. The options -A and -Z can nevertheless be used to force words to appear in the first (respectively last) position of the displayed line.

Note that the number of requested columns will be automatically reduced if a word does not fit in the calculated column size.

In this mode each column has the same width.

By default, the spaces surrounding the output string will be deleted. This option forces them to be retained.
By default, when searching, an alarm is produced by the terminal when the user enters a character or makes a move which lead to no result or to an error condition. This argument make this beep visual by briefly showing the cursor.
-s pattern
Place the cursor on the first word corresponding to the specified pattern.

pattern can be:

A # immediately followed by a number giving the initial position of the cursor (counting from 0).

If the word at this position is excluded, then the first previous non excluded word is selected if it exists, otherwise the first non excluded word is selected.

If this number if greater than the number of words, the cursor will be set on the latest selectable position.

A single # or the string #last to set the initial cursor position to the latest selectable word position.
A string starting with a / indicating that we want the cursor to be set to the first word matching the given regular expression.
A prefix string indicating that we want the cursor to be set on the first word matching the string given (a will match Cancel by example).

Warning, when searching for a prefix or a regular expression, smenu only looks for them after an eventual modification, so for example, the command: smenu -I/c/x/ -s/c <<< "a b c d" won't find c and put the cursor on a but smenu -I/c/x/v -s/c <<< "a b c d" will find it and put the cursor on the x substituting the c on screen only

\u sequences can be used in the pattern.

-m message
Displays a message above the window. If the current locale is not UTF-8, then all UTF-8 characters in it will be converted into a dot.

\u sequences can be used in the message.

Note that the message will be truncated if it does not fit on a terminal line.

When -t is followed by a number of columns, the default is to compact the columns so that they use the less terminal width as possible. This option enlarges the columns in order to use the whole terminal width.

When in column mode, -w can be used to force all the columns to have the same size (the largest one). See option -c below.

Note that the column's size is only calculated once when the words are displayed for the first time. A terminal resize will not update this value. This choice enables a faster display.

Tells the program to clean up the display before quitting by removing the selection window after use as if it was never displayed.
Centers the display if possible.
Sets the column mode. In this mode the lines of words do not wrap when the right border of the terminal is reached but only when a special character is read. Some words will not be displayed without an horizontal scrolling.

If such a scrolling is needed, some indications may appear on the left and right edge of the window to help the user to reach the unseen words.

In this mode, the width of each column is minimal to keep the maximum information visible on the terminal.

Sets the line mode. This mode is the same as column mode but without any column alignment.
Enables ENTER to validate the selection even in search mode.
Replaces all non-printable characters by a blank. If this results in a blank word, it will be potentially deleted.
Sets the display attributes of the elements displayed and the cursor.

At least one attribute prefixed attribute must be given.

PREFIX can take the following values:

included words.
excluded words.
scroll bar.
shift indicator.
tagged words.
cursor on tagged words.
search field.
search text.
approximate search field with error.
approximate search text with error.
matching words field.
matching words text.
matching words field with error.
matching words text with error.
direct access tag.

If more than one attribute is given, then they must be separated by spaces.

See the -1 option for the ATTR syntax.

-i regex
Sets the include filter to match the selectable words. All the other words will become implicitly non-selectable (excluded)

-i can be used more than once with cumulative effect.

\u sequences can also be used in the regexp.

-e regex
Sets the exclude filter to match the non-selectable words. All the other selectable words will become implicitly selectable (included)

-e can be used more than once with cumulative effect. This filter has a higher priority than the include filter.

The regex selections made using -i and/or -e are done before the possible words alterations made by -I or -E (see below).

\u sequences can also be used in the regexp.

-C [i|e] <col selectors>

These letters are case independent so I can be used in place of i per example.

In column mode, this option allows one to restrict the previous selections or de-selections to some columns. If no selection is given via -i and -e this option gives the possibility to select entire columns by giving their numbers (1 based) of extended regular expressions.

i or nothing select the specified ranges of columns. e select all but the specified ranges of columns.

The words in the selected columns will be considered as included And the others excluded.

A selection by regular expressions means that a column containing a word matching one of these expression will be included or excluded according to the letter given after the option.

Regular expressions and column numbers can be freely mixed.

Regular expression in -C and -R can contain UTF-8 characters either directly or by using the \u notation.

Example of columns selection: -Ci2,3,/X./,5-7 forces the cursor to only navigate in columns 2,3,5,6 and 7 and those containing a two characters word starting with 'X'. If e was used in place of i, all the columns would have been selected except the columns 2,3,5,6,7 and those matching the extended regular expression 'X.'.

Spaces are allowed in the selection string if they are protected.

The column mode is forced when this option is selected.

-R [i|e] <row selectors>
Similar to -C but for the rows.

One difference though: this is the line mode which is forced by this option NOT the column mode.

-C and -R can be used more than once in a cumulative manner: The selection mode (selection or de-selection) is given by the first occurrence of the options, the other occurrences will only update the selected or de-selected ranges.

-S /regex/replacement string/[g][v][s]
Post-processes the words by applying a regular expression based substitution. The argument must be formatted as in the sed editor.

This option can be used more than once. Each substitution will be applied in sequence on each word. This sequence can be stopped if a stop flag is encountered.

The optional trailing g (for global) means that all matching occurrences shall be replaced and not only the first one.
The optional trailing v (for visual) means that the altered words will only be used for display and search. The modifications will not be reflected in the returned word.
The optional trailing s (for stop) means that no more substitution will be allowed on this word even if another -S is used.
The optional trailing i (for ignore case) means that the string search operation should ignore the case for this pattern.

Small example: R=$(echo a b c | smenu -S /b/B/) will display "a B c" and R will contain B if B is selected meanwhile R=$(echo a b c | smenu -S /b/B/v) will display the same as above but R will contain the original word b if B is selected. In both cases, only the word B will be searchable and not b.

-I /regex/replacement string/[g][v][s]
Post-processes the selectable words by applying a regular expression based substitution (see -S for details).
-E /regex/replacement string/[g][v][s]
Post-processes the excluded (or non-selectable) words by applying a regular expression based substitution (see -S for details).

The / separator that -I and -E are using above can be substituted by any other character except SPACE, \t, \f, \n, \r and \v.

In the three previous options, regex is a POSIX Extended Regular Expression. For details, please refer to the regex manual page.

Additionally \u sequences can also be used in the regexp.

If a post-processing action (-S/-I/-E) results in an empty (length 0) word, then we have two cases:
in column mode:
Substitutions involving empty words can lead to misalignments, so it is necessary to prohibit them and terminate the program. These substitutions have to be made with other tools before using this utility.
The word is simply removed.
-A regex
In column mode, forces all words matching the given regular expression to be the first one in the displayed line. If you want to only rely on this method to build the lines, just specify an empty regex to set the end-of-line separator with -L '')

\u sequences can also be used in the regexp after -A.
-Z regex
Similar to -A but forces the word to be the latest of its line. The same trick with -L can also be used.

\u sequences can also be used in the regexp after -Z.
-N [regex]
This option allows one to number the selectable words matching a specific regular expression. These numbers are numbered starting from 1 and provides a direct access to the words.

To use this functionality, the user must enter the number which corresponds to the desired entry digit per digit.

Each new digit must be added in a time frame of 1/2 seconds (per default) otherwise the number is considered complete and a newly entered digit will start a new number. If the number does not exists, then the cursor is restored to it's initial position.

The sub-options of the -D option described below can change the way -N sets and formats the numbers.

This option can be used more than once with cumulative effects.

-N, -U and -F can be mixed.

-U [regex]
This option allows one to un-number words. If placed after a previous -N, it can be used to remove the numbering of selected words. If placed before, the word which doesn't match its regular expression will be numbered by default.

This mechanism is similar to to the inclusion/exclusion of words by -i and -e.

This option can be used more than once with cumulative effects.

-U, -N and -F can be mixed.

This option is similar to -N but does not generate a continuous flow of numbers but extracts them from the word itself.

With this option you can take full control of the numbering of the displayed word. Note that the numbering does not need to be ordered.

The resulting word after the extraction of the number must be non empty.

Some sub-option are required, see the -D option described below.

Notice that for this option to work correctly, all the embedded numbers must have the same number of digits. To get that, a preprocessing may be necessary on the words before using this program.

-F, -N and -U can be mixed.

-D [parameters]
This option allows one to change the default behaviour of the -N, -U and -F options.

Its optional parameters are called sub-options and must respect the format x:y where x can be:

l (-F, -N and -U options)
Here y is the UTF-8 character (in native or \u form) to print before the number. The default is a single space.
r (-F, -N and -U options)
Here y is the UTF-8 character (in native or \u form) to print after the number. The default is ).
a (-F, -N and -U options)
Here y is 'left' (or one of its prefixes) if the number must be left aligned, or 'right' (or one of its prefixes) if it must be right aligned. The default is right.
p (-F, -N and -U options)
Here y is 'included' (or one of its prefixes) or 'all' (or one of its prefixes) for the initial padding of the non numbered words. 'included' means that only included word will be padded while 'all' means pad all words. The default is all.
w (-F, -N and -U options)
Here y is the width of the number between 1 and 5 included.
f (-F, -N and -U options)
Here y controls if the numbering must follow the last extracted number (defaults to yes) or if it must remain independent.
h (-F option)
Tells what to do with the characters present before the embedded number if any.

The allowed directives are: 'trim' which discads them if they form an empty word (only made of spaces and tabulations), 'cut' which unconditionnaly discards them and 'keep' which places them at the beginning of the resulting word.

The default value for this directive is 'keep'.

o (-F option)
Here y is the offset of the first multibyte character of the number to extract from the word (defaults to 0).
n (-F option)
Here y is the number of multibyte characters to extract from the word starting at the offset given by the o sub-option.
i (-F option)
Here y is number of multibyte characters to ignore after the extracted number
d (-F, -N and -U options)
Here y is a multibyte separator. When present, this directive instructs smenu to output the selected numbered word(s) prefixed by its(their) direct access number(s) and the given separator.

Only the numbered word(s) will be prefixed.

d stands for decorate.

This directive can be useful when you want to post-process the output according to its direct access number.

s (-F, -N and -U options)
Here y is the direct access number that will be set for the first numbered word. Its value is 1 by default, a value of 0 is possible.

Example: r:\> l:\< a:l d:_

To number all words with the default parameters, use the syntax: "-N ." which is a shortcut for: "-N . l:' ' r:')' a:r p:a"

The padding sub-option specifies whether spaces must also be added in front of excluded words or not to improve compactness.

When the w sub-option is not given the width of the numbers is determined automatically but if -F is set and the value of the n sub-option is given then this value is used.

-1 ... -5 regex [ATTR]
Allows one to give a special display color to up to 5 classes of words specified by regular expressions. They are called special levels. Only selectable words will be considered.

By default, the 5 special levels have their foreground color set to red, green, brown/yellow, purple and cyan. All these colors also can be set or modified permanently in the configuration files. See the example file above for an example.

The optional second argument (ATTR) can be used to override the default or configured attributes of each class. Its syntax is the same as the one used in the configuration file:

[fg][/bg][,{b|d|r|s|u|i}] | [{b|d|r|s|u|i}]

Examples of possible attributes are:

  2/0,bu green on black bold underline
  /2     green background
  5      text in purple
  rb     reverse bold

\u sequences can be used in the pattern.

-g [string]
Replaces the blank after each words in column or tabular mode by a column separator.

This separator is extracted from the string argument and each of its (multibyte) character is used one after the other to fill the gutter.

If there are more columns that gutter characters then the last character is used for the remaining columns.

When not given, the separator defaults to a vertical bar | (or a full height vertical bar if the locale is set to UTF-8).

Each character can be given in normal or \u form in the string argument.

Example: "|- " will allow one to separate the first two columns with '|', then '-' will be used and ' ' will separate the remaining columns if any.

Prevents the display of the scroll bar.
-W bytes
This option can be used to specify the characters (or multibyte sequences) which will be used to delimit the input words.

Multibyte sequences (UTF-8) can be natives of using the same ASCII representation used in words (a leading \u following by up to 8 hexadecimal characters).

Non-printable characters in arguments should be given using the standard $'' representation. $'\t' stands for the tabulation character for example.

The default delimiters are: SPACE, $'\t' and $'\n'.

-L bytes
This option can be used to specify the characters (or multibyte sequences) which will be used to delimit the lines in the input stream.

Multibyte sequences (UTF-8) can be natives of using the same ASCII representation used in words (a leading \u following by up to 8 hexadecimal characters).

Non-printable characters in arguments should be given using the standard $'' representation. $'\n' stands for the newline character for example.

The default delimiter is: $'\n'.

This option is only useful when the -c or -l option is also set.

The characters (or multibyte sequences) passed to -L are automatically added to the list of word delimiters as if -W was also used.

\u sequences can also be used here.

-T [separator]
Enables the multi-selections or tag mode. In this mode, several selectable words can be selected without leaving the program.

The current word can be automatically tagged when the ENTER key is pressed to complete the selection process if the -p option is also set or if no word has been tagged.

All the tagged words (and possibly the world under the cursor) will be sent to stdout separated by the optional argument given after the option -T.

Note than this separator can have more than one character, contain UTF-8 characters (in native or \u form) and can even contain control character as in $'\n'.

A space is used as the default separator if none is given.

Caution: To get exactly the same behavior as in version 0.9.11 and earlier, you must also use the -p option.

-P [separator]
Works like -T but, unlike -T, the output depends on the order in which the words were tagged. In other words, the first tagged word comes first in the output, the second tagged word comes next, and so on. -P stands for "Pin".
This option modifies the default behavior of the -T and -P options. An untagged word under the cursor will be automatically tagged when ENTER is pressed.
Displays the current version and quits.
-x type [word] delay
-X type [word] delay Sets a timeout. Three types of timeout are possible:
At the timeout, the word under the cursor and/or the tagged words are sent to the standard output if the ENTER key has been pressed
At the timeout, nothing is selected as if the q key has been pressed
At the timeout, the word given after the type is selected. Note that this word doesn't need to be part of the words coming from the standard input.

Each type can be be shortened as a prefix of its full name ("cur" for "current" of "q" for "quit" per example).

The delay must be set in seconds and cannot be above 99999 seconds.

The remaining time (in seconds) is added at the end of the message displayed above the selection window and is updated in real time each second.

Each key press except ENTER, q, Q and ^C resets the timer to its initial value.

The -X version works like -x but no periodic remaining messages is displayed above the selection window.

-/ search_method
Affects the '/' key to a search method. By default '/' is affected to 'fuzzy' but the argument can be any prefix of 'prefix', 'substring' or 'fuzzy'.


If tabulators (\t) are embedded in the input, there is no way to replace them with the original number of spaces. In this case use another filter (like expand) to pre-process the data.  




Simple Yes/No/Cancel request with "No" as default choice:

In bash:
  read R <<< $(echo "Yes No Cancel" \
               | smenu  -d -m "Please choose:" -s /N)

  R=$(echo "Yes No Cancel" \
      | smenu -d -m "Please choose:" -s /N)

In ksh:
  print "Yes No Cancel"                \
  | smenu -d -m "Please choose:" -s /N \
  | read R



Get a 3 columns report about VM statistics for the current process in bash/ksh on Linux:

R=$(grep Vm /proc/$$/status | expand | smenu -b -W$'\n' -t3 -g -d)



Create a one column selection window containing the list of the first 20 LVM physical volumes. At the end, the selection window will be erased. This example is written in ksh).

pvs -a -o pv_name --noheadings                 \
| smenu -m "PV list" -n20 -t1 -d -s //dev/root \
| read R

The display will have a look similar to the following with the cursor set on the word /dev/root:

PV list
/dev/md126           \
/dev/md127           |
/dev/root            | <- cursor here.
/dev/sda2            |
/dev/sdb2            |
/dev/sdc1            |
/dev/sdc2            |
/dev/system/homevol  /


4 (advanced)

Imagine a file named sample.mnu with the following content:

"1 First Entry" "3 Third entry"
"2 Second entry" "4 Fourth entry"
@@@ "5 Fifth entry"
"0 Exit menu"

Then this quite esoteric command will render it (centered on the screen) as:

|            Test menu             |
|                                  |
| 1) First Entry   3) Third entry  |
| 2) Second entry  4) Fourth entry |
|                  5) Fifth entry  |
|                                  |
| 0) Exit menu                     |

with the cursor on Quit and only the numbers and "Quit" selectable.

R=$(smenu R=$(./smenu -q -d -s/Exit -M -n 30 -c \
                      -e "@+" -E '/@+/ /'            \
                      -F -D n:1 i:1                  \
                      -m "Test menu"$' < sample.mnu)

The selected entry will be available in R

Try to understand it as an exercise.  


NO_COLOR: force a monochrome terminal when set.  


Some terminal emulators, those notably based on VTE version later than 0.35 (see, have a new feature that gives them the possibility to wrap/unwrap already displayed lines when resizing the window.

As far as I known, there is no terminfo entry to disable that.

On these types of terminals, the automatic re-display of the output of smenu will be disturbed and some artifacts may appear on the screen if the terminal window is resized.  


© 2015 Pierre Gentile (



Moving among words
Searching for a word
Selection and Exit
Tagging (multi-selections)
Scroll bar
Terminal resizing (also see BUGS/LIMITATIONS)
Unicode support
4 (advanced)

This document was created by