kak@hico2.UUCP (Kris A. Kugel) (03/16/91)
Am I confused, or does egrep on the 3b1
handle multiple regular expressions incorrectly?
egrep 'foo|bar'
seems to match '[fb][oa][or]' patterns,
instead of what I wanted (matching lines with "foo" or "bar")
Kris A. Kugel
( 908 ) 842-2707
uunet!tsdiag.ccur.com!hico2!kak
{daver,ditka,zorch}!hico2!kak
internet: kak@hico2.westmark.comdavid@twg.com (David S. Herron) (03/21/91)
In article <1268@hico2.UUCP> kak@hico2.UUCP (Kris A. Kugel) writes: >Am I confused, or does egrep on the 3b1 >handle multiple regular expressions incorrectly? > > egrep 'foo|bar' > >seems to match '[fb][oa][or]' patterns, >instead of what I wanted (matching lines with "foo" or "bar") Try egrep '(foo)|(bar)' And read the F'ing manual a little more closely next time ... -- <- David Herron, an MMDF & WIN/MHS guy, <david@twg.com> <- Formerly: David Herron -- NonResident E-Mail Hack <david@ms.uky.edu> <- <- "MS-DOS? Where we're going we don't need MS-DOS." --Back To The Future
kak@hico2.UUCP (Kris A. Kugel) (03/22/91)
> <- David Herron, an MMDF & WIN/MHS guy, <david@twg.com> In article <8783@gollum.twg.com>, david@twg.com (David S. Herron) writes: > In article <1268@hico2.UUCP> kak@hico2.UUCP (Kris A. Kugel) writes: > >Am I confused, or does egrep on the 3b1 > >handle multiple regular expressions incorrectly? > > > > egrep 'foo|bar' > > > >seems to match '[fb][oa][or]' patterns, > >instead of what I wanted (matching lines with "foo" or "bar") > > Try > > egrep '(foo)|(bar)' I'll try this, thanks for the method. > And read the F'ing manual a little more closely next time ... No need to get hostile, I did after your message. Carefully. It says, "3. Two regular expressions separated by | or by a new line match strings that are matched by either. 4. A regular expression may be enclosed in parentheses () for grouping. The order of precedence of operators is [], then * ? + , then concatenation, then | and new-line." Now, if you want to complain that I missed something obvious and well documented, fine. If what you describe is really how the feature is designed, then I claim that that is NOT obvious from the description on the manual page. Now, I could see from this description that maybe the pattern I tried would match "fo[bo]ar", but that was NOT the behavior I remember seeing. I claim that either the manual page is very poorly written, or else what I tried should have given me CLEARLY different results than I got. Now, to find the egrep on osu-cis . . . . Kris A. Kugel ( 908 ) 842-2707 uunet!tsdiag.ccur.com!hico2!kak {daver,ditka,zorch}!hico2!kak internet: kak@hico2.westmark.com
andrew@alice.att.com (Andrew Hume) (03/22/91)
In article <8783@gollum.twg.com>, david@twg.com (David S. Herron) writes: ~ In article <1268@hico2.UUCP> kak@hico2.UUCP (Kris A. Kugel) writes: ~ >Am I confused, or does egrep on the 3b1 ~ >handle multiple regular expressions incorrectly? ~ > ~ > egrep 'foo|bar' ~ > ~ >seems to match '[fb][oa][or]' patterns, ~ >instead of what I wanted (matching lines with "foo" or "bar") ~ ~ Try ~ ~ egrep '(foo)|(bar)' ~ ~ And read the F'ing manual a little more closely next time ... ~ ~ ~ -- ~ <- David Herron, an MMDF & WIN/MHS guy, <david@twg.com> ~ <- Formerly: David Herron -- NonResident E-Mail Hack <david@ms.uky.edu> ~ <- ~ <- "MS-DOS? Where we're going we don't need MS-DOS." --Back To The Future with respect, mr herron, kris apparently did read the manual and his/her expression is perfectly valid. the parentheses in your retort are superfluous. if after typing egrep 'foo|bar' foo bar far <ctrl-d> you get only the lines foo bar echoed back, then all is well. if anything different happens, your egrep is buggered. andrew@research.att.com
guest@geech.ai.mit.edu (Guest Account) (03/23/91)
Since without any kind of quoting or special chars a regular expression is 1 character: egrep 'foo|bar' matches lines containing fooar and fobar. That's why you need to use parentheses. You have to read the manual page for ed(1) to get the rest of the story on regular expressions. To quote the grep man page: "Egrep accepts regular expressions as in ed(1), except..." Daniel Guilderson ryan@cs.umb.edu
tkacik@rphroy.ph.gmr.com (Tom Tkacik) (03/23/91)
In article <GUEST.91Mar22115440@geech.ai.mit.edu>, guest@geech.ai.mit.edu (Guest Account) writes: |> Since without any kind of quoting or special chars a regular |> expression is 1 character: |> egrep 'foo|bar' |> matches lines containing fooar and fobar. That's why you need to use |> parentheses. You have to read the manual page for ed(1) to get the |> rest of the story on regular expressions. To quote the grep man page: Sorry, try again. Concatenation has higher precedence than '|'. egrep 'foo|bar' will match either foo or bar, not fobar nor fooar. egrep works like that here at work. I have not yet tried it at home. I will. If it works as Kris A. Kugel says, then it must be busted. -- Tom Tkacik tkacik@clyde.cs.gmr.com GM Research Labs tkacik@kyzyl.mi.org "I'm president of the United States, and I'm not going to eat anymore broccoli." --- George Bush
guy@cbnewsc.att.com (guy.r.berentsen) (03/23/91)
> |> egrep 'foo|bar' > |> matches lines containing fooar and fobar. That's why you need to use > > egrep 'foo|bar' will match either foo or bar, not fobar nor fooar. > On my unix pc at work (running 3.5) egrep 'foo|bar' matches foo or bar (of course it also matches "fobar" and "fooar" since the later contains the string "foo" and the former contains the string "bar")
tkacik@kyzyl.mi.org (Tom Tkacik) (03/23/91)
In article <1268@hico2.UUCP>, kak@hico2.UUCP (Kris A. Kugel) writes: > Am I confused, or does egrep on the 3b1 > handle multiple regular expressions incorrectly? > > egrep 'foo|bar' > > seems to match '[fb][oa][or]' patterns, > instead of what I wanted (matching lines with "foo" or "bar") According to the man page egrep 'foo|bar' should match any line containing either 'foo' or 'bar'. I just tried it on kyzyl, and it seems to work properly. It does not match any line containing either 'boo' or 'far', (or other incorrect pattern mentioned above). I suggest that Kris check again. Someone stated that the | has higher precedence than concatination, and that the above is equivalent to egrep 'fo(o|b)ar', and will match either 'fooar' or 'fobar'. It will match these, but only because they are a subset of the patterns it will match. -- Tom Tkacik | tkacik@kyzyl.mi.org | To rent this space, call 1-800-555-QUIP. ...!rphroy!kyzyl!tkacik |
kak@hico2.UUCP (Kris A. Kugel) (03/23/91)
> > In article <1268@hico2.UUCP> kak@hico2.UUCP (Kris A. Kugel) writes: > > >Am I confused, or does egrep on the 3b1 > > >handle multiple regular expressions incorrectly? > > > > > > egrep 'foo|bar' > > > > > >seems to match '[fb][oa][or]' patterns, > > Now, I could see from this [manual] description that maybe the > pattern I tried would match "fo[bo]ar", but that was NOT the > behavior I remember seeing. Well, evidently the problem is not as simple as I described it. I'd used the "foo|bar" example as a substitute phrase for what I was really trying. Bruce Lilly tried this literally with a good sample file, and with this simple case egrep seems to work correctly. And his test works correctly on my machine, too. If I recall correctly, I saw the (I say) incorrect behavior when I trying to find a newsarticle about some subject (I didn't remember the title or group exactly, so I had an "egrep -i "pat1|pat2|pat3|pat4|pat5|pat6|pat7|pat9" kind of thing, probably running on stdin.) I saw the malfunction, looked at a couple of (false match) lines, and thought "these could be explained by . . ." without doing a comprehensive check. It sounds like I'm going to have to find a test case showing this problem exactly, because it's CLEARLY not the straightforward malfunction that I described in my original article. (*sigh* when I get time . . . .) I'd also conclude that I should have been more accurate and complete about the conditions under which I saw the problem, and presented the "[fb][oa][or]" explanation as a hypothesis, rather than as the observed problem. I'll try to be more accurate in the future. Kris A. Kugel ( 908 ) 842-2707 uunet!tsdiag.ccur.com!hico2!kak {daver,ditka,zorch}!hico2!kak internet: kak@hico2.westmark.com