Shell Programming and Scripting

Give it a try with:-

Code:

tr " " "\n" < your_log_file_name | grep -i "^another$"

Does this get you what you need?

I suppose you might have to consider another., another,, another!, etc. too.

If this is a worry, try:-

Code:

tr "[:punct:]" " " < your_log_file_name | grep -i "^another$"

The plan with this one is to translate all punctuation to spaces, then translate all spaces to a new-line, then use grep to count the records (one word each by now) that contain the string from the first to the last character only.


It might be worth testing it out a small section first and think of as many variations as you can think of.



Robin

Your use of split() is redundant. AWK already split the line. The value that split() returns is the current value of NF. You can iterate through the fields using i<=NF and checking each $i.

Also, if there aren't any matches, count will be undefined and the END print statement will output an empty line. I would change its argument to count+0.

Beware of matching substrings triggering false positives.


That's seems to me to be a reasonable approach, but neither of those pipelines actually implements it. As described, the approach would require a pipeline with two tr's. However, it can be done with one if you convert punctuation directly to newlines, which would be equivalent. In that case, you can modify your latter suggestion to:

Code:

Code:

tr '[:punct:] ' '[\n*]' < your_log_file_name | grep -i "^another$"

Note the space after the punctuation character class. If one wanted to include any blank characters, the :blank: class could have been used instead.

Often times it's easier and safer to define what to include than what to exclude. Based on your approach, if we define a word as a sequence of [:alpha:] characters, the following portable solution can be used:

Code:

tr -sc '[:alpha:]' '[\n*]' | grep -Fixc word

Regards,
Alister

Thanks for pointing out my logical error.

Perhaps I should have gone with:-

Code:

tr "[:punct:]" " " < your_log_file_name | tr "[:blank:]" "\n" | grep -i "^another$"

.
.
I am a little confused by your suggestion to use the :alpha: class. Would this not act on the characters we want to preserve?

I got some odd output from a quick test:-

Code:

# echo "Hello world!" | tr "[:alpha:]" "\n*"
***** *****!
# echo "Hello world!" | tr "[:alpha:]" "\n"
echo "Hello world!" | tr "[:alpha:]" "\n" 





 




!

Am I missing something?

I could bunch it up into a single tr too. A quick test shows this:-

Code:

# echo "This is what I am, I am not spam I hope." | tr "[:punct:][:blank:]" "\n"|grep -c "^am$"
2

That would consolidate my suggestion to:-

Code:

tr "[:punct:][:blank:]" "\n" < your_log_file_name | grep -i "^another$"

Robin

Two things. First, the -c option, which complements [:alpha:], so what is matched is everything that is not a member of [:alpha:].

Second, \n* must be bracketed. With the brackets, it represents as many newlines as it takes to match the length of the class in the previous argument. Without the brackets, it means a single newline followed by as many asterisks as it takes. So, your erroneous version would replace the first character in [:alpha:] with a newline and every subsequent character with an asterisk.

In my current locale, A is the first member of [:alpha:]. Note how in the first example 'A' is converted to a newline while 'a' becomes an asterisk:

Code:

$ printf aAa | tr '[:alpha:]' '\n*' | od -c
0000000   *  \n   *
0000003
$ printf aAa | tr '[:alpha:]' '[\n*]' | od -c
0000000  \n  \n  \n
0000003

With modern tr implementations, you can probably get away with simply using \n:

Code:

$ printf aAa | tr '[:alpha:]' '\n' | od -c
0000000  \n  \n  \n
0000003

... but there is the possibility of a portability issue. From POSIX tr:
Regards,
Alister

Hai alister,
can u please say me the exact command that i can use to find the particular word 'another' from 17000 lines from a log file or text file

Again, Does it really take hard to try at least one solution yourself?
If you can not access your system now, please come back to us whenever you try.

Thanks for ur advice