Sed Regular Expressions: Difference between revisions

From NovaOrdis Knowledge Base
Jump to navigation Jump to search
 
(20 intermediate revisions by the same user not shown)
Line 1: Line 1:
=External=
* https://www.gnu.org/software/sed/manual/html_node/Regular-Expressions.html
=Internal=
=Internal=


Line 4: Line 6:
* [[Regular Expressions]]
* [[Regular Expressions]]


=Special Characters (need to be escaped in regular expressions)=
=<span id='Special_Characters_.28need_to_be_escaped_in_regular_expressions.29'></span>Meta-Characters - Special Characters (need to be escaped in regular expressions)=
 
<font size=-2>
  /
  /
  "
  "
Line 15: Line 17:
  :
  :
  * # zero or more
  * # zero or more
+ # one or more - does not seem to work for all seds.
  . # dot
  . # dot
</font>
Single quote is a special case, to match it use its ASCII hexadecimal value prefixed by \x as follows, instead of escaping it:
<font size=-2>
\x27
</font>
To use () for grouping, they need to be escaped:
<font size=-2>
\(...\)
</font>
More details in [[#Grouping|Grouping]] below.


=Non-Special Characters (do not need to be escaped in regular expressions)=
=Non-Special Characters (do not need to be escaped in regular expressions)=
 
<font size=-2>
  <
  <
  >
  >
Line 28: Line 39:
  {
  {
  }
  }
+ # this is interesting, I thought '+' is a meta-character, more experimentation necessary.
</font>
 
==<tt>+</tt>==
 
<code>+</code> by itself is not a meta-character, it matches "+".
 
The GNU version matches <code>\+</code> with "one or more characters"


=Grouping=
=Grouping=
Line 34: Line 51:
Use <tt>\(</tt> and <tt>\)</tt> for grouping. Parentheses must be escaped to be interpreted as grouping separator.
Use <tt>\(</tt> and <tt>\)</tt> for grouping. Parentheses must be escaped to be interpreted as grouping separator.


=Brackets=
Brackets mean "any one of"
<font size=-2>
[ab]
</font>
will match "a" or "b".
<syntaxhighlight lang='bash'>
echo "blah" | sed -s 's/[ab]/x/g'
</syntaxhighlight>
prints "xlxh".
==Brackets and Negation==
Match everything except the specified characters. More than one characters is matched. The behavior is different from the [[Bash_Patterns#Negation|behavior of bash patterns on negation]]:
<font size=-2>
[^abc]*
</font>
=Match Zero or One Character=
Normally, this would be achieved with <code>?</code> placed after the character or the group of characters in questions, but this not work with standard <code>sed</code>.


=Examples=
=Examples=
Line 39: Line 80:
Match everything except space:
Match everything except space:


<pre>
<font size=-2>
    [^ ]*
[^ ]*
</pre>
</font>


<pre>
<font size=-2>
    .*
.*
</pre>
</font>


seems to work too.
seems to work too.
Line 51: Line 92:
Words (digits, alpha, _):
Words (digits, alpha, _):


<pre>
<font size=-2>
sed -e 's/[0-9a-zA-Z_]*/THIS_WAS_A_WORD/g'
sed -e 's/[0-9a-zA-Z_]*/THIS_WAS_A_WORD/g'
</pre>
</font>


Blank spaces (spaces, tabs, newlines): <tt>\s</tt> does not seem to work.
Blank spaces (spaces, tabs, newlines): <tt>\s</tt> does not seem to work.


=Regular Expression Syntax=
=Regular Expression Syntax=
<font color=red>TO NORMALIZE across [[Java_Regular_Expressions#Regular_Expression_Syntax|java Regular Expression Syntax]], [[Grep_Regular_Expressions#Regular_Expression_Syntax|grep Regular Expression Syntax]], [[Sed_Regular_Expressions#Regular_Expression_Syntax|sed Regular Expression Syntax]].</font>
<font color=darkkhaki>TO NORMALIZE across [[Java_Regular_Expressions#Regular_Expression_Syntax|java Regular Expression Syntax]], [[Grep_Regular_Expressions#Regular_Expression_Syntax|grep Regular Expression Syntax]], [[Sed_Regular_Expressions#Regular_Expression_Syntax|sed Regular Expression Syntax]].</font>

Latest revision as of 21:46, 1 April 2024

External

Internal

Meta-Characters - Special Characters (need to be escaped in regular expressions)

/
"
$ # unescaped signifies end of line 
^ # unescaped signifies the beginning of a line
!
[
]
:
* # zero or more
. # dot

Single quote is a special case, to match it use its ASCII hexadecimal value prefixed by \x as follows, instead of escaping it:

\x27

To use () for grouping, they need to be escaped:

\(...\)

More details in Grouping below.

Non-Special Characters (do not need to be escaped in regular expressions)

<
>
(
)
!
-
{
}

+

+ by itself is not a meta-character, it matches "+".

The GNU version matches \+ with "one or more characters"

Grouping

Use \( and \) for grouping. Parentheses must be escaped to be interpreted as grouping separator.

Brackets

Brackets mean "any one of"

[ab]

will match "a" or "b".

echo "blah" | sed -s 's/[ab]/x/g'

prints "xlxh".

Brackets and Negation

Match everything except the specified characters. More than one characters is matched. The behavior is different from the behavior of bash patterns on negation:

[^abc]*

Match Zero or One Character

Normally, this would be achieved with ? placed after the character or the group of characters in questions, but this not work with standard sed.

Examples

Match everything except space:

[^ ]*

.*

seems to work too.

Words (digits, alpha, _):

sed -e 's/[0-9a-zA-Z_]*/THIS_WAS_A_WORD/g'

Blank spaces (spaces, tabs, newlines): \s does not seem to work.

Regular Expression Syntax

TO NORMALIZE across java Regular Expression Syntax, grep Regular Expression Syntax, sed Regular Expression Syntax.