bash
— standard shell
A thorough understanding of bash
programming is vital when working with
ebuilds.
Bash conditionals
Multiple selection
Multiple pronged selection can be done using else
and elif
:
if something ; then
do_stuff
elif something_else ; then
do_other_stuff
elif full_moon ; then
howl
else
turn_into_a_newt
fi
if some_stuff ; then
# A statement is required here. a blank or a comment
# isn't enough!
else
einfo "Not some stuff"
fi
If you really don't want to restructure the block, you can use a single colon
(:
) on its own as a null statement.
if some_stuff ; then
# Do nothing
:
else
einfo "Not some stuff"
fi
Selection tests
To do comparisons or file attribute tests, [[ ]]
(preferred) or
[ ]
blocks are needed.
# is ${foo} zero length?
if [[ -z ${foo} ]] ; then
die "Please set foo"
fi
# is ${foo} equal to "moo"?
if [[ ${foo} == "moo" ]] ; then
einfo "Hello Larry"
fi
# does ${ROOT}/etc/deleteme exist?
if [[ -f ${ROOT}/etc/deleteme ]] ; then
einfo "Please delete ${ROOT}/etc/deleteme manually!"
fi
Single versus double brackets in bash
[[ ]]
form is generally safer than [ ]
and should be used in
all new code.
POSIX compliance is not a concern for ebuilds, as their interpreter is guaranteed to be GNU Bash. POSIX style tests have different semantics and using the common forms of tests adheres to the principle of least surprise. Most developers will be used to Bash test semantics and behaviour and deviating from this in ebuilds may be confusing.
This is because [[ ]]
is a bash syntax construct, whereas [ ]
is a
program which happens to be implemented as an internal — as such, cleaner
syntax is possible with the former. For a simple illustration, consider:
bash$ [ -n ${foo} ] && [ -z ${foo} ] && echo "huh?"
huh?
bash$ [[ -n ${foo} ]] && [[ -z ${foo} ]] && echo "huh?"
bash$
String comparison in bash
The general form of a string comparison is string1 operator string2
. The
following are available:
Operator | Purpose |
---|---|
== (also = ) |
String equality |
!= |
String inequality |
< |
String lexicographic comparison (before) |
> |
String lexicographic comparison (after) |
=~ |
String regular expression match |
String tests in bash
The general form of string tests is -operator "string"
. The following are
available:
Operator | Purpose |
---|---|
-z "string" |
String has zero length |
-n "string" |
String has non-zero length |
Integer comparison in bash
The general form of integer comparisons is int1 -operator int2
. The
following are available:
Operator | Purpose |
---|---|
-eq |
Integer equality |
-ne |
Integer inequality |
-lt |
Integer less than |
-le |
Integer less than or equal to |
-gt |
Integer greater than |
-ge |
Integer greater than or equal to |
File tests in bash
The general form of a file test is -operator "filename"
. The following are
available (lifted from man bash
):
Operator | Purpose |
---|---|
-a file |
Exists (use -e instead) |
-b file |
Exists and is a block special file |
-c file |
Exists and is a character special file |
-d file |
Exists and is a directory |
-e file |
Exists |
-f file |
Exists and is a regular file |
-g file |
Exists and is set-group-id |
-h file |
Exists and is a symbolic link |
-k file |
Exists and its sticky bit is set |
-p file |
Exists and is a named pipe (FIFO) |
-r file |
Exists and is readable |
-s file |
Exists and has a size greater than zero |
-t fd |
Descriptor fd is open and refers to a terminal |
-u file |
Exists and its set-user-id bit is set |
-w file |
Exists and is writable |
-x file |
Exists and is executable |
-O file |
Exists and is owned by the effective user id |
-G file |
Exists and is owned by the effective group id |
-L file |
Exists and is a symbolic link |
-S file |
Exists and is a socket |
-N file |
Exists and has been modified since it was last read |
File comparison in bash
The general form of a file comparison is "file1" -operator "file2"
.
The following are available:
Operator | Purpose |
---|---|
file1 -nt file2 |
file1 is newer (according to modification date) than file2, or file1 exists and file2 does not |
file1 -ot file2 |
file1 is older than file2, or file2 exists and file1 does not |
file1 -ef file2 |
file1 is a hard link to file2 |
Boolean algebra in bash
There are constructs available for boolean algebra ('and', 'or' and 'not').
These are used outside of the [[ ]]
blocks. For operator precedence, use
( )
.
Construct | Effect |
---|---|
first || second |
first or second (short circuit) |
first && second |
first and second (short circuit) |
! condition |
not condition |
[[ ]]
constructs, and using
!
before a test is fairly common. [[ ! -f foo ]] && bar
is fine. However, there are catches — [[ -f foo && bar ]]
will not work properly, since commands cannot be run inside [[ ]]
blocks.
Inside [ ]
blocks, several -test
style boolean operators are available.
These should be avoided in favour of [[ ]]
and the above operators.
Bash iterative structures
There are a few simple iterative structures available from within bash
. The
most useful of these is a for
loop. This can be used to perform the same
task upon multiple items.
for myvar in "the first" "the second" "and the third" ; do
einfo "This is ${myvar}"
done
There is a second form of the for
loop which can be used for repeating an
event a given number of times.
for (( i = 1 ; i <= 10 ; i++ )) ; do
einfo "i is ${i}"
done
There is also a while
loop, although this is usually not useful within
ebuilds.
while hungry ; do
eat_cookies
done
This is most commonly used to iterate over lines in a file:
while read myline ; do
einfo "It says ${myline}"
done < some_file
See die and subshells
for an explanation of why while read < file
should
be used over cat file | while read
.
Bash variable manipulation
There are a number of special ${}
constructs in bash
which either
manipulate or return information based upon variables. These can be used instead
of expensive (or illegal, if we're in global scope) external calls to sed
and friends.
bash
string length
The ${#somevar}
construct can be used to obtain the length of a string
variable.
somevar="Hello World"
echo "${somevar} is ${#somevar} characters long"
bash
variable default values
There are a number of ways of using a default value if a variable is unset or
zero length. The ${var:-value}
construct expands to the value of ${var}
if it is set and not null, or value
otherwise. The ${var-value}
construct is similar, but checks only that the variable is set.
The ${var:=value}
and ${var=value}
forms will also assign value
to
var
if var
is unset (and also set, but null for the :=
form).
The ${var:?message}
form will display message
to stderr and then exit if
var
is unset or null. This should generally not be used within ebuilds as it
does not use the die
mechanism. There is a ${var?message}
form too.
The ${var:+value}
form expands to value
if var
is set and not null,
or a blank string otherwise. There is a ${var+value}
form.
bash
substring extraction
The ${var:offset}
and ${var:offset:length}
constructs can be used to
obtain a substring. Strings are zero-indexed. Both offset
and length
are
arithmetic expressions.
The first form with a positive offset returns a substring starting with the
character at offset
and continuing to the end of a string. If the offset is
negative, the offset is taken relative to the end of the string.
${var:0-1}
. ${var:-1}
will
not work.
The second form returns the first length
characters of the value of
${var}
starting at offset
. If offset
is negative, the offset is
taken from the end of the string. The length
parameter must not be less
than zero. Again, negative offset
values must be given as an expression.
bash
command substitution
The $(command)
construct can be used to run a command and capture the
output (stdout
) as a string.
`command`
construct also does this, but should be avoided in
favour of $(command)
for clarity, ease of reading and nesting purposes.
myconf="$(use_enable acl) $(use_enable nls) --with-tlib=ncurses"
bash
string replacements
There are three basic string replacement forms available: ${var#pattern}
,
${var%pattern}
and ${var/pattern/replacement}
. The first two are used
for deleting content from the start and end of a string respectively. The third
is used to replace a match with different content.
The ${var#pattern}
form will return var
with the shortest match of
pattern
at the start of the value of var
deleted. If no match can be
made, the value of var
is given. To delete the longest match at the start,
use ${var##pattern}
instead.
The ${var%pattern}
and ${var%%pattern}
forms are similar, but delete the
shortest and longest matches at the end of var
respectively.
%
and
#
being the non-greedy forms). This is arguably incorrect, but the terms
are fairly close.
The ${var/pattern/replacement}
construct expands to the value of var
with the first match of pattern
replaced with replacement
. To replace
all matches, ${var//pattern/replacement}
can be used.
man bash
incorrectly describes what will be matched. Of all the possible
leftmost matches, the longest will be taken. Yes, really, the longest, even if
it involves favouring later groups or later branches. This is not like
perl
or sed
.
See
IEEE Std 1003.1-2017, section 9.1 for details.
To match only if pattern
occurs at the start of the value of var
, the
pattern should begin with a #
character. To match only at the end, the
pattern should begin with a %
.
If replacement
is null, matches are deleted and the /
following
pattern
may be omitted.
The pattern
may contain a number of special metacharacters for pattern
matching.
Character | Meaning |
---|---|
* |
Matches any string, including the null string |
? |
Matches any single character |
[...] |
Matches any one of the enclosed characters |
Refer to the Bash Reference Manual for further details and caveats regarding these characters.
If the extglob
shell option is enabled, a number of additional constructs
are available. These can be extremely useful sometimes. In the following
table, a pattern-list
is a list of one or more patterns separated by
|
.
Construct | Meaning |
---|---|
?(pattern-list) |
Matches zero or one occurrence of the given patterns |
*(pattern-list) |
Matches zero or more occurrences of the given patterns |
+(pattern-list) |
Matches one or more occurrences of the given patterns |
@(pattern-list) |
Matches one of the given patterns |
!(pattern-list) |
Matches anything except one of the given patterns |
bash
arithmetic expansion
The $(( expression ))
construct can be used for integer arithmetic
evaluation. expression
is a C-like arithmetic expression. The following
operators are supported (the table is in order of precedence, highest first):
Operators | Effect |
---|---|
var++ , var--
|
Variable post-increment, post-decrement |
++var , --var
|
Variable pre-increment, pre-decrement |
- , +
|
Unary minus and plus |
! , ~
|
Logical negation, bitwise negation |
** |
Exponentiation |
* , / , %
|
Multiplication, division, remainder |
+ , -
|
Addition, subtraction |
<< , >>
|
Left, right bitwise shifts |
<= , >= , < , >
|
Comparison: less than or equal to, greater than or equal to, strictly less than, strictly greater than |
== , !=
|
Equality, inequality |
& |
Bitwise AND |
^ |
Bitwise exclusive OR |
| |
Bitwise OR |
&& |
Logical AND |
|| |
Logical OR |
expr ? expr : expr |
Conditional operator |
= , *= , /= , %= , += , -= ,
<<= , >>= , &= , ^= , |=
|
Assignment |
expr1 , expr2 |
Multiple statements |
**=
assignment operator.