Atualização:
Os melhores seriam:
sed 's/^.\+\(\/\|\&\|\?\)v=\([^\&]*\).*//'
awk 'match($0,/((\/|&|\?)v=)([^&]*)/,x){print x[3]}'
grep -Po '(?<=(\/|&|\?)v=)[^&]*'
# Saying match / or & then v=
RFC 3986 afirma:
URI = scheme ":" hier-part [ "?" query ] [ "#" fragment ]
query = *( pchar / "/" / "?" )
fragment = *( pchar / "/" / "?" )
pchar = unreserved / pct-encoded / sub-delims / ":" / "@"
unreserved = ALPHA / DIGIT / "-" / "." / "_" / "~"
sub-delims = "!" / "$" / "&" / "'" / "(" / ")"
/ "*" / "+" / "," / ";" / "="
…
Então, para ser um uso seguro:
| sed 's/#.*//' | - to remove #fragment part
na frente.
Ou seja,
| sed 's/#.*//' | grep -Po '(?<=(\/|&)v=)[^&]*'
SED (2):
echo 'http://www.youtube.com/watch?v=qdRaf3-OEh4&playnext=1&list=PL4367CEDBC117AEC6&feature=results_main' \
| sed 's/^.\+\Wv=\([^\&]*\).*//'
Explicação:
's
/…/…/ /THIS/WITH THIS/
'substitute/MATCH 0 or MORE THINGS and GROUP them in ()/WITH THIS/
+-------------------------- s _s_ubsititute
|+------------------------- / START MATCH
|| +---- / END MATCH
|| | +-- REPLACE WITH - ==Group 1. Or FIRS low ().
|| | | +- / End of SUBSTITUTE
s/^.\+\Wv=\([^\&]*\).*//'
+++-+-+-+-+-----+-+------- ^ Match from beginning of line
++-+-+-+-+-----+-+------- . Match any character
+-+-+-+-+-----+-+------- \+ multiple times (grep (greedy +, * *? etc))
+-+-+-+-----+-+------- \W Non-word-character
+-+-+-----+-+------- v= Literally match "v="
+-+-----+-+------- \( Start MATCH GROUP
+-----+-+------- [^\&]* Match any character BUT & - as many as possible
+-+------- \) End MATCH GROUP
+------- .* Match anything; *As many times as possible
- aka to end of line; as there is no
[abc] would match a OR b OR c
[abc]* would match a AND/OR b AND/OR c - as many times as possible
[^abc] would match anything BUT a,b or c
// Replace ENTIRE match with MATCH GROUP number 1.
That would be - everything between \( and \) - which his anything but "&"
after the literal string "v=" - which in turn has a non word letter in
front of it.
That also means that no match means no substitution which ultimately result in
no change.
Resultado: qdRaf3-OEh4
Nota: Se nenhuma sequência inteira for retornada.
(G) AWK:
echo 'http://www.youtube.com/watch?v=qdRaf3-OEh4&playnext=1&list=PL4367CEDBC117AEC6&feature=results_main' \
| awk 'match($0,/(\Wv=)([^&]*)/,v){print v[2]}'
Resultado: qdRaf3-OEh4
Explicação:
Em Awk match(string, regexp) é uma função que procura a correspondência mais longa e mais à esquerda de regexp em string. Aqui eu usei uma extensão que vem com o Gawk. (veja Awk , GAwk ; MAwk etc.) que coloca as correspondências individuais - isto é: o que está entre parênteses - em uma matriz de correspondências.
O padrão é bem parecido com o do Perl / Grep abaixo.
+-------------------------------------- Built in function
| +--------------------------------- Entire input ($1 would have been filed 1)
| | etc. (Using default delimiters " "*)
| |
| |
| | (....)(....) ------------------ Places \Wv= in one group 1, and [^&]* group 2.
match($0, /(\Wv=)([^&]*)/, v){print v[2]}
| | | |
| | +-+---- Use "v" from /, v; v is a user defined name
| | +---- 2 specifies index in v, which is group from
| | what is between ()'s in /…/
| |
| +----------- Print is another built in function.
+--------------- Group name that one can use in print.
GREP (usando Perl-compatível):
echo 'http://www.youtube.com/watch?v=qdRaf3-OEh4&playnext=1&list=PL4367CEDBC117AEC6&feature=results_main' | \
grep -Po '(?<=\Wv=)[^&]*'
Resultado: qdRaf3-OEh4
Explicação:
-P Use Perl compatible
-o Only print match of the expression.
- That means: Of our pattern only print/return what it matches.
If nothing matches; return nothing.
+------- ^ Negate math to - do not match (ONLY as it is FIRST between [])
|+------ & A literal "&" character
||
(?<=\Wv=)[^&]*
| | | | ||
| | | | |+---- * Greedy; as many times as possible.
| | | +--+----- [] Wild order/any order of what is inside []
| | +----------- v= Literal v=
| +------------- \W Non Word character
+----------------- (?<= What follows should be (mediately) preceded by.
?=Huh, <=left, = =Equals to
So: Match literal "v=" where "v" is preceded by an non-word-character. Then match
anything; as many times as possible until we are at end of line or we meet an "&".
As you can't have "&" in an URL between key/value pairs this should be OK.