fast alternative to grep file multiple times? - bash

I currently use long piped bash commands to extract data from text files like this, where $f is my file:
result=$(grep "entry t $t " $f | cut -d ' ' -f 5,19 | \
sort -nk2 | tail -n 1 | cut -d ' ' -f 1)
I use a script that might do hundreds of similar searches of $f ,sorting selected lines in various ways depending on what I'm pulling out. I like one-line bash strings with a bunch of pipes because its compact and easy, but it can take forever. Can anyone suggest a faster alternative? Maybe something that loads the whole file into memory first?
Thanks

You might get a boost with doing the whole pipe with gawk or another awk that has asorti by doing:
contents="$(cat "$f")"
result="$(awk -vpattern="entry t $t" '$0 ~ pattern {matches[$5]=$19} END {asorti(matches,inds); print inds[1]}' <<<"$contents")"
This will read "$f" into a variable then we'll use a single awk command (well, gawk anyway) to do all the rest of the work. Here's how that works:
-vpattern="entry t $t": defines an awk variable named pattern that contains the shell variable t
$0 ~ pattern matches the current line against the pattern, if it matches we'll do the part in the braces, otherwise we skip it
matches[$5]=$19 adds an entry to an array (and creates the array if needed) where the key is the 5th field and the value is the 19th
END do the following function after all the input has been processed
asorti(matches,inds) sort the entries of matches such that the inds is an array holding the order of the keys in matches to get the values in sorted order
print inds[1] prints the index in matches (i.e., a $5 from before) associated with the lowest 19th field
<<<"$contents" have awk work on the value in the shell variable contents as though it were a file it was reading
Then you can just update the pattern for each, not have to read the file from disk each time and not need so many extra processes for all the pipes.
You'll have to benchmark to see if it's really faster or not though, and if performance is important you really should think about moving to a "proper" language instead of shell scripting.

Since you haven't provided sample input/output this is just a guess and I only post it because there's other answers already posted that you should not do, so - this may be what you want instead of that one line:
result=$(awk -v t="$t" '
BEGIN { regexp = "entry t " t " " }
$0 ~ regexp {
if ( ($6 > maxKey) || (maxKey == "") ) {
maxKey = $6
maxVal = $5
}
}
END { print maxVal }
' "$f")
I suspect your real performance issue, however, isn't that script but that you are running it and maybe others inside a loop that you haven't shown us. If so, see why-is-using-a-shell-loop-to-process-text-considered-bad-practice and post a better example so we can help you.

Related

How can I generate multiple counts from a file without re-reading it multiple times?

I have large files of HTTP access logs and I'm trying to generate hourly counts for a specific query string. Obviously, the correct solution is to dump everything into splunk or graylog or something, but I can't set all that up at the moment for this one-time deal.
The quick-and-dirty is:
for hour in 0{0..9} {10..23}
do
grep $QUERY $FILE | egrep -c "^\S* $hour:"
# or, alternately
# egrep -c "^\S* $hour:.*$QUERY" $FILE
# not sure which one's better
done
But these files average 15-20M lines, and I really don't want to parse through each file 24 times. It would be far more efficient to parse the file and count each instance of $hour in one go. Is there any way to accomplish this?
You can ask grep to output the matching part of each line with -o and then use uniq -c to count the results:
grep "$QUERY" "$FILE" | grep -o "^\S* [0-2][0-9]:" | sed 's/^\S* //' | uniq -c
The sed command is there to keep only the two digit hour and the colon, which you can also remove with another sed expression if you want.
Caveats: this solution works with GNU grep and GNU sed, and will produce no output, rather than "0", for hours with no log entries. Kudos to #EdMorton for pointing these issues out in the comments, and other issues that were fixed in the answer above.
Assuming the timestamp appears with a space before the 2-digit hour, then a colon after
gawk -v patt="$QUERY" '
$0 ~ patt && match($0, / ([0-9][0-9]):/, m) {
print > (m[1] "." FILENAME)
}
' "$FILE"
This will create 24 files.
Requires GNU awk for the 3-arg form of match()
This is probably what you really need, using GNU awk for the 3rd arg to match() and making assumptions about what your input might look like, what your QUERY variable might contain, and what the output should look like:
awk -v query="$QUERY" '
match($0, " ([0-9][0-9]):.*"query, a) { cnt[a[1]+0]++ }
END {
for (hr=0; hr<=23; hr++) {
printf "%02d = %d\n", hr, cnt[hr]
}
}
' "$FILE"
Don't really use all upper case for non-exported shell variables btw - see Correct Bash and shell script variable capitalization.

Utilising variables in tail command

I am trying to export characters from a reference file in which their byte position is known. To do this, I have a long list of numbers stored as a variable which have been used as the input to a tail command.
For example, the reference file looks like:
ggaaatgcattcaaacatgc
And the list looks like:
5
10
7
15
I have tried using this code:
list=$(<pos.txt)
echo "$list"
cat ref.txt | tail -c +"list" | head -c1 > out.txt
However, it keeps returning "invalid number of bytes: '+5\n10\n7\n15...'"
My expected output would be
a
t
g
a
...
Can anybody tell me what I'm doing wrong? Thanks!
It looks like you are trying to access your list variable in your tail command. You can access it like this: $list rather than just using quotes around it.
Your logic is flawed even after fixing the variable access. The list variable includes all lines of your list.txt file. Including the newline character \n which is invisible in many UIs and programs, but it is of course visible when you are manually reading single bytes. You need to feed the lines one by one to make it work properly.
Also unless those numbers are indexes from the end, you need to feed them to head instead of tail.
If I understood what you are attempting to do correctly, this should work:
while read line
do
head -c $line ref.txt | tail -c 1 >> out.txt
done < pos.txt
The reason for your command failure is simple. The variable list contains a multi-line string stored from the pos.txt files including newlines. You cannot pass not more than one integer value for the -c flag.
Your attempts can be fixed quite easily with removing calls to cat and using a temporary variable to hold the file content
while IFS= read -r lineNo; do
tail -c "$lineNo" ref.txt | head -c1
done < pos.txt
But then if your intentions is print the desired output in a new-line every time, head does not output that way. It just forms a string atga for your given input in a single line and not across multiple lines with one character at each line.
As Gordon mentions in one of the comments, for much more efficient FASTA files processing, you could just use one invocation of awk though (skipping multiple forks to head/tail). Your provided input does not involve any headers to skip which would be straightforward as
awk ' FNR==NR{ n = split($0,arr,""); for(i=1;i<=n;i++) hash[i] = arr[i] }
( $0 in hash ){ print hash[$0] } ' ref.txt pos.txt
You could use cut instead of tail:
pos=$(<pos.txt)
cut -c ${pos//$'\n'/,} --output-delimiter=$'\n' ref.txt
Or just awk:
awk -F '' 'NR==FNR{c[$0];next} {for(i in c) print $i}' pos.txt ref.txt
both yield:
a
g
t
a

Batch create files with name and content based on input file

I am a mac OS user trying to batch create a bunch of files. I have a text file with column of several hundred terms/subjects, eg:
hydrogen
oxygen
nitrogen
carbon
etcetera
I want to programmatically fill a directory with text files generated from this subject list. For example, "hydrogen.txt" and "oxygen.txt" and so on, with each file created by iterating through the lines of my list_of_names.txt file. Some lines are one word, but other lines are two or three words (eg: "carbon monoxide"). This I have figured out how to do:
awk 'NF>0' list_of_names.txt | while read line; do touch "${line}.txt"; done
Additionally I need to create two lines of content within each of these files, and the content is both static and dynamic...
# filename
#elements/filename
...where in the example above the pound sign ("#") and "elements/" would be the same in all of the files created, but "filename" would be variable (eg: "hydrogen" for "hydrogen.txt" and "oxygen" for "oxygen.txt" etc). One further wrinkle is that if any spaces appear at all on the second line of content, there needs to be a trailing pound sign. For example:
# filename
#elements/carbon monoxide#
...although this last part is not a dealbreaker and I can use grep to modify list_of_names.txt such that phrases like "carbon monoxide" become "carbon_monoxide" and just deal with the repercussions of this later. (But if it is easy to preserve the spaces, I would prefer that.)
After a couple hours of searching and attempts to use sed, awk, and so on I am stuck at a directory full of files with the correct filename.txt format, but I can't get further that this. Mostly I think my efforts are failing because the solutions I can find for doing something like this are using commands I am not familiar with and they are structured for GNU and don't execute correctly in Terminal on Mac OS.
I am amenable to processing this in multiple steps (ie make all of the files.txt first, then run a second step to populate the content of the files), or as a single command that makes the files and all of their content simultaneously ('simultaneously' from a human timescale).
My horrible pseudocode (IN CAPS) for how this would look as 2 steps:
awk 'NF>0' list_of_names.txt | while read line; do touch "${line}.txt"; done
awk 'NF>0' list_of_names.txt | while read line; OPEN "${line}.txt" AND PRINT "# ${line}\n#elements/${line}"; IF ${line} CONTAINS CHARACTER " " PRINT "#"; done
You could use a simple Bash loop and create the files in one shot:
#!/bin/bash
while read -r name; do # loop through input file content
[[ $name ]] || continue # skip empty lines
output=("# $name") # initialize the array with first element
trailing=
[[ $name = *" "* ]] && trailing="#" # name has spaces in it
output+=("#elements/$name$trailing") # name doesn't have a space
printf '%s\n' "${output[#]}" > "$name.txt" # write array content to the output file
done < list_of_names.txt
Doing it in awk:
awk '
NF {
trailing = (/ / ? "#" : "")
out=$0".txt"
printf("# %s\n#elements/%s%s\n", $0, $0, trailing) > out
close(out)
}
' list_of_names.txt
Doing the whole job in awk will yield better performance than in bash, which isn't really suited to processing text like this.
It seems to me that this should cover the requirements you've specified:
awk '
{
out=$0 ".txt"
printf "# %s\n#elements/%s%s\n", $0, $0, (/ / ? "#" : "") >> out
close(out)
}
' list_of_subjects.txt
Though you could shrink it to a one-liner:
awk '{printf "# %s\n# elements/%s%s\n",$0,$0,(/ /?"#":"")>($0".txt");close($0".txt")}' list_of_subjects.txt

How to find integer values and compare them then transfer the main files?

I have some output files (5000 files) of .log which are the results of QM computations. Inside each file there are two special lines indicate the number of electrons and orbitals, like this below as an example (with exact spaces as in output files):
Number of electrons = 9
Number of orbitals = 13
I thought about a script (bash or Fortran), as a solution to this problem, which grep these two lines (at same time) and get the corresponding integer values (9 and 13, for instance), compare them and finds the difference between two values, and finally, list them in a new text file with the corresponding filenames.
I would really appreciate any help given.
Am posting an attempt in GNU Awk, and have tested it in that only.
#!/bin/bash
for file in *.log
do
awk -F'=[[:blank:]]*' '/Number of/{printf "%s%s",$2,(NR%2?" ":RS)}' "$file" | awk 'function abs(v) {return v < 0 ? -v : v} {print abs($1-$2)}' >> output_"$file"
done
The reason I split the AWK logic to two was to reduce the complexity in doing it in single huge command. The first part is for extracting the numbers from your log file in a columnar format and second for getting their absolute value.
I will break-down the AWK logic:-
-F'=[[:blank:]]*' is a mult0 character delimiter logic including = and one or more instances of [[:blank:]] whitespace characters.
'/Number of/{printf "%s%s",$2,(NR%2?" ":RS)}' searches for lines starting with Number of and prints it in a columnar fashion, i.e. as 9 13 from your sample file.
The second part is self-explanatory. I have written a function to get the absolute value from the two returned values and print it.
Each output is saved in a file named output_, for you to process it further.
Run the script from your command line as bash script.sh, where script.sh is the name containing the above lines.
Update:-
In case if you are interested in negative values too i.e. without the absolute function, change the awk statement to
awk -F'=[[:blank:]]*' '/Number of/{printf "%s%s",$2,(NR%2?" ":RS)}' "$file" | awk '{print ($1-$2)}' >> output_"$file"
Bad way to do it (but it will work)-
while read file
do
first=$(awk -F= '/^Number/ {print $2}' "$file" | head -1)
second=$(awk -F= '/^Number/ {print $2}' "$file" | tail -1)
if [ "$first" -gt "$second" ]
then
echo $(("$first" - "$second"))
else
echo $(("$second" - "$first"))
fi > "$file"_answer ;
done < list_of_files
This method picks up the values (in the awk one liner and compares them.
It then subtracts them to give you one value which it saves in the file called "$file"_answer. i.e. the initial file name with '_answer' as a suffix to the name.
You may need to tweak this code to fit your purposes exactly.

Grep outputs multiple lines, need while loop

I have a script which uses grep to find lines in a text file (ics calendar to be specific)
My script finds a date match, then goes up and down a few lines to copy the summary and start time of the appointment into a separate variable. The problem I have is that I'm going to have multiple appointments at the same time, and I need to run through the whole process for each result in grep.
Example:
LINE=`grep -F -n 20130304T232200 /path/to/calendar.ics | cut -f1 d:`
And it outputs only the lines, such as
86 89
Then it goes on to capture my other variables, as such:
SUMMARYLINE=$(( $LINE + 5 ))
SUMMARY:`sed -n "$SUMMARYLINE"p /path/to/calendar.ics
my script runs fine with one output, but it obviously won't work with more than 1 and I need for it to. should I send the grep results into an array? a separate text file to read from? I'm sure I'll need a while loop in here somehow. Need some help please.
You can call grep from a loop quite easily:
while IFS=':' read -r LINE notused # avoids the use of cut
do
# First field is now in $LINE
# Further processing
done < <(grep -F -n 20130304T232200 /path/to/calendar.ics)
However, if the file is not too large then it might be easier to read the whole file into an array and more around that.
With your proposed solution, you are reading through the file several times. Using awk, you can do it in one pass:
awk -F: -v time=20130304T232200 '
$1 == "SUMMARY" {summary = substr($0,9)}
/^DTSTART/ {start = $2}
/^END:VEVENT/ && start == time {print summary}
' calendar.ics

Resources