(old)Puppy Linux Discussion Forum

READ-ONLY Archive
https://oldforum.puppylinux.com/

Bash: Count numbers in list

https://oldforum.puppylinux.com/viewtopic.php?t=80491

Page 1 of 2

Posted: Fri 24 Aug 2012, 14:47
by zigbert
This sounds interesting.
In am on the run this weekend, so I can't compare executing-time until Monday.


Sigmund

Posted: Fri 24 Aug 2012, 14:50
by zigbert
technosaurus wrote:Oops need to prepend the -lt test with a [ $1 ] && ... For the case where it has not been played in the time period
....and that means???? :oops:

Posted: Fri 24 Aug 2012, 16:27
by jamesbond
My submission:

Code: Select all

#!/bin/ash
LANG=C

FIND=3421  # timestamp to find (random)
TIMES=3650 # 3650 timestamps (tracks played everyday for 10 years) - number from 1 to $TIMES
SONGS=2000 # 2000 songs

# generate fake timestamp data
tstamps=$(seq -s, 1 $TIMES)
line="/path/to/song.mp3 | $tstamps"
#echo $line

# generate fake playlist (repeat timestamp data $SONGS times)
for a in $(seq 1 $SONGS); do
	echo $line
done | 

# do actual work
awk -F, -v TS=$FIND '{
		split($1, a, "|"); $1=a[2] # fix first entry
		max=NF; min=1; 
		
		while (max-min > 1) {
			i=int( (max+min)/2 )
			if (TS >= $i) min = i
			else max = i
		}
		print a[1], min
}'
With TIMES=3650 and SONGS=2000 as above (meaning, 3650 timestamps per song, with 2000 songs), result:

Code: Select all

# time ./fav.sh  > /dev/null

real	0m1.641s
user	0m2.916s
sys	0m0.073s
.

Using Ziggy's original parameter of 25 timestamps per song and 2000 songs (TIMES=25, SONGS=2000), result

Code: Select all

# time ./fav.sh  > /dev/null

real	0m0.068s
user	0m0.080s
sys	0m0.010s
Using 25 timestamps for 20,000 songs (TIMES=25, SONGS=20000), result:

Code: Select all

# time ./fav.sh  > /dev/null

real	0m0.479s
user	0m0.730s
sys	0m0.057s
Using 3650 timestamps for 20,000 songs (TIMES=3650, SONGS=20000), result:

Code: Select all

# time ./fav.sh  > /dev/null

real	0m16.855s
user	0m30.308s
sys	0m0.520s
Of course these are on my machine and is not directly comparable with anyone else's number. Ziggy need to run this on his own machine to test :)

cheers!

Count numbers in list

Posted: Fri 24 Aug 2012, 17:05
by L18L
zigbert wrote:
technosaurus wrote:Oops need to prepend the -lt test with a [ $1 ] && ... For the case where it has not been played in the time period
....and that means???? :oops:
change

Code: Select all

while ([ $1 -lt $timestamp ]) do
to

Code: Select all

while ( [ $1 ] && [ $1 -lt $timestamp ]) do
tested successfully (no more: ash: 1345623045: unknown operand ):
# ./file_times_played_since 1345648694
The number of times played /root/file1.mp3 since 1345648694 is 0
The number of times played /root/file2.mp3 since 1345648694 is 1
#

Re: Count numbers in list

Posted: Fri 24 Aug 2012, 17:16
by technosaurus
@jamesbond - over 100 lines grep becomes faster than using builtins, and it doesn't slow it down significantly for less than 100 lines, so its probably worth calling grep in this case since we can't assume a user has/plays only a few songs.
L18L wrote:
zigbert wrote:
technosaurus wrote:Oops need to prepend the -lt test with a [ $1 ] && ... For the case where it has not been played in the time period
....and that means???? :oops:
change

Code: Select all

while ([ $1 -lt $timestamp ]) do
to

Code: Select all

while ( [ $1 ] && [ $1 -lt $timestamp ]) do
tested successfully (no more: ash: 1345623045: unknown operand ):
# ./file_times_played_since 1345648694
The number of times played /root/file1.mp3 since 1345648694 is 0
The number of times played /root/file2.mp3 since 1345648694 is 1
#
thanks, I was/am posting from my phone.

Posted: Fri 24 Aug 2012, 17:55
by akash_rawal
jamesbond wrote: My submission:
...
I tested your solution, it takes only one second to process 100 songs with 100000 timestamps. Faster by orders of magnitude :)

Count numbers in list

Posted: Sat 25 Aug 2012, 10:45
by L18L
I have been testing jamesbond´s submission, too.

change
print a[1], min
to
print a[1], NF-min
to get TIMES from FIND to NOW as required.

Speed: wow, it is TIME to learn awk now. 8)

Posted: Sat 25 Aug 2012, 12:32
by technosaurus
@L18L
the awk would be even faster if it only computed stuff for the match:

Code: Select all

FILEPATH=${FILEPATH//\//\\\/} #escape the slashes for awk ... if you care about paths
#FILEPATH=${FILEPATH##*/} #remove the path for awk ... if you only care about names
awk -F, -v TS=$FIND '/'$FILEPATH'/{...
not sure how well awk works on non-ascii filenames though

Count numbers in list

Posted: Sun 26 Aug 2012, 07:56
by L18L
technosaurus wrote:the awk would be even faster if it only computed stuff for the match
Yes, and slower if we care about accuracy.

The binary search algorithm is not accurate if case of some identic values.
But on a single user system there are no doubles :lol:

Posted: Sun 26 Aug 2012, 18:56
by zigbert
Test results so far with the following file

Code: Select all

/root/file1.mp3| 1345641102,1345644563,1345647584,1345647585,1345647586,1345647587,1345647588,1345647589,1345647590,1345740286,1345740287,1345740288,1345740290,1345740291,1345740291,1345740292,1345740293,1345740293,1345740294,1345740295,1345740296,1345740297,1345740298,1345740299,1345740300
/root/file2.mp3| 1345623045,1345643786,1345648695,1345648696,1345648697,1345648698,1345648699,1345648700,1345648701,1345740286,1345740287,1345740288,1345740290,1345740291,1345740291,1345740292,1345740293,1345740293,1345740294,1345740295,1345740296,1345740297,1345740298,1345740299,1345740300
[...]
/root/file1999.mp3| 1345641102,1345644563,1345647584,1345647585,1345647586,1345647587,1345647588,1345647589,1345647590,1345740286,1345740287,1345740288,1345740290,1345740291,1345740291,1345740292,1345740293,1345740293,1345740294,1345740295,1345740296,1345740297,1345740298,1345740299,1345740300
/root/file2000øæå.mp3| 1345623045,1345643786,1345648695,1345648696,1345648697,1345648698,1345648699,1345648700,1345648701,1345740286,1345740287,1345740288,1345740290,1345740291,1345740291,1345740292,1345740293,1345740293,1345740294,1345740295,1345740296,1345740297,1345740298,1345740299,1345740300
########################################

Code: Select all

#!/bin/bash

#usage: played_since_foreach time

time="$1";

ifs_bak="$IFS"
IFS=""

while read -d "|" record; do
   IFS=","
   #Get all timestamps in array
   read -a times
   #Binary search
   i=$(( ${#times[*]}/2 ))
   d=$(( $i/2 ))
   test "$d" -eq "0" && d=1 #precaution
   ulim=$(( ${#times[*]}-1 ))
   count=-1
   while test "$i" -ge "0" -a "$i" -lt "$ulim"; do #check whether index is in range
      if test "$time" -lt "${times[$i]}"; then
         i=$(( $i-$d ))
      elif test "$time" -gt "${times[$(( $i+1 ))]}"; then
         i=$(( $i+$d ))
      else
         #found, tell no. of records greater than this
         count="$(( $ulim-$i ))"
         break
      fi
      d=$(( ($d/2) ))
      test "$d" -eq "0" && d=1 #precaution
   done
   #In case of 'out of range'
   if test "$count" -lt "0"; then
      if test "$ulim" -lt "0"; then
         count=0
      else
         if test "$time" -lt "$times"; then
            count="${#times[*]}"
         elif test "$time" -gt "${times[$ulim]}"; then
            count=0
         else
            count="bug" #This should not be reached
         fi
      fi
   fi
   #Show result for this record
   echo "$record: $count"
done < database.txt

IFS="$ifs_bak"

Code: Select all

real	0m1.129s
user	0m1.080s
sys	0m0.047s
########################################

Code: Select all

#!/bin/ash
timestamp=$1

ifs_bak=$IFS
IFS=","

while read LOGENTRY; do

filename=${LOGENTRY%%|*}
#LOGENTRY=`grep $filename database.txt`
#grep is faster than while read case on large files

STAMPS=${LOGENTRY#*|}
#you could use bash arrays instead - set allows "arrays" in sh, ash, etc...
#set ${STAMPS//,/ }
set ${STAMPS}
#for speed change IFS to "," vs. the //,/ } but don't forget to reset it
while ( [ $1 ] && [ $1 -lt $timestamp ]) do
shift #a hacky way to remove all entries < timestamp
done
# $# is the number of args passed or in this case set with "set ..."
echo The number of times played $filename since $timestamp is $#

done < database.txt

IFS=$ifs_bak

Code: Select all

real	0m1.434s
user	0m0.190s
sys	0m0.633s
########################################

Code: Select all

#!/bin/ash
LANG=C
FIND=$1

cat /root/database.txt | 
awk -F, -v TS=$FIND '{
      split($1, a, "|"); $1=a[2] # fix first entry
      max=NF; min=1;
      
      while (max-min > 1) {
         i=int( (max+min)/2 )
         if (TS >= $i) min = i
         else max = i
      }
      print a[1], NF-min 
}'

Code: Select all

real	0m0.026s
user	0m0.020s
sys	0m0.003s
########################################

All worked with non-english filename /root/file2000øæå.mp3


Thank you all
Sigmund

Posted: Sun 26 Aug 2012, 19:30
by zigbert
Ok, next stage
I don't understand a bit of the awk-code, so I need some help to integrate it fully into pMusic code.

It now works ok (only ok) to find the last played songs for a given period:
My NEW index file looks right now like this:

Code: Select all

  /mnt/sdb1/musikk/mp3/Judas priest - Monsters of rock.mp3|    Judas priest - Monsters of rock.mp3|:,1345762237
  /mnt/sdb1/musikk/mp3/Iron maiden - Seventh son of a seventh son.mp3|    Iron maiden - Seventh son of a seventh son.mp3|:,1345762833
  /mnt/sdb1/musikk/mp3/Savatage - Dead winter dead.mp3|    Savatage - Dead winter dead.mp3|:,1345763094
  /mnt/sdb1/musikk/mp3/Twisted sister - Hot love.mp3|    Twisted sister - Hot love.mp3|:,1345763569
  /mnt/sdb1/musikk/mp3/Twisted sister - I believe in you.mp3|    Twisted sister - I believe in you.mp3|:,1345763892
  /mnt/sdb1/musikk/mp3/Twisted sister - I'm so hot for you.mp3|    Twisted sister - I'm so hot for you.mp3|:,1345764140
  /mnt/sdb1/musikk/mp3/Twisted sister - I wanna rock.mp3|    Twisted sister - I wanna rock.mp3|:,1345764326
  /mnt/sdb1/musikk/mp3/Twisted sister - King of the fools.mp3|    Twisted sister - King of the fools.mp3|:,1345764715
  /mnt/sdb1/musikk/mp3/Twisted sister - Leader of the pack.mp3|    Twisted sister - Leader of the pack.mp3|:,1345764940
  /mnt/sdb1/musikk/mp3/Twisted sister - Lookin out for no. 1.mp3|    Twisted sister - Lookin out for no. 1.mp3|:,1345765130
  /mnt/sdb1/musikk/mp3/Twisted sister - One bad habit.mp3|    Twisted sister - One bad habit.mp3|:,1345765538
  /mnt/sdb1/musikk/mp3/Twisted sister - Run for your life.mp3|    Twisted sister - Run for your life.mp3|:,1345766019
  /mnt/sdb1/musikk/mp3/Twisted sister - The beast.mp3|    Twisted sister - The beast.mp3|:,1345760757,1345766230
  /mnt/sdb1/musikk/mp3/Twisted sister - The price.mp3|    Twisted sister - The price.mp3|:,1345766462
  /mnt/sdb1/musikk/mp3/Twisted sister - Under the blade.mp3|    Twisted sister - Under the blade.mp3|:,1345766745
  /mnt/sdb1/musikk/mp3/Twisted sister - Wake up (the sleeping giant).mp3|    Twisted sister - Wake up (the sleeping giant).mp3|:,1345767008
  /mnt/sdb1/musikk/mp3/Twisted sister - We're not gonna take it.mp3|    Twisted sister - We're not gonna take it.mp3|:,1345767230
  /mnt/sdb1/musikk/mp3/Twisted sister - You can't stop rock'n'roll.mp3|    Twisted sister - You can't stop rock'n'roll.mp3|:,1345767514
  /mnt/sdb1/musikk/mp3/Twisted sister - You're not alone (Suzette's song).mp3|    Twisted sister - You're not alone (Suzette's song).mp3|:,1345767762
  /mnt/sdb1/musikk/mp3/Twisted sister - You want what we got.mp3|    Twisted sister - You want what we got.mp3|:,1345767984
  /mnt/sdb1/musikk/mp3/Saxon - Wheels of steel.mp3|    Saxon - Wheels of steel.mp3|:,1345789853
  /mnt/sdb1/musikk/mp3/Judas priest - Parental guidance (LP).mp3|    Judas priest - Parental guidance (LP).mp3|:,1345790062
  /mnt/sdb1/musikk/mp3/Judas priest - Turbo lover (LP).mp3|    Judas priest - Turbo lover (LP).mp3|:,1345790660
  /mnt/sdb1/musikk/mp3/Alice Cooper - Poison.mp3|    Alice Cooper - Poison.mp3|:,1345790933
  /mnt/sdb1/musikk/mp3/Faith no more - Easy.mp3|    Faith no more - Easy.mp3|:,1345791129
  /mnt/sdb1/musikk/mp3/Bangles - Manic monday.mp3|    Bangles - Manic monday.mp3|:,1345791315
  /mnt/sdb1/musikk/mp3/Secret garden - Nocturne.mp3|    Secret garden - Nocturne.mp3|:,1345791511
  /mnt/sdb1/musikk/mp3/Dimmu borgir - Vredesbyrd.mp3|    Dimmu borgir - Vredesbyrd.mp3|:,1345760201,1345760237,1345794966
  /mnt/sdb1/musikk/mp3/Black sabbath - Heaven and hell.mp3|    Black sabbath - Heaven and hell.mp3|:,1345760180,1345761905,1345796352
  /mnt/sdb1/musikk/mp3/Twisted sister - Love is for suckers.mp3|    Twisted sister - Love is for suckers.mp3|:,1345765338,1345797921
  /mnt/sdb1/musikk/mp3/Dimmu borgir - Burn in hell.mp3|    Dimmu borgir - Burn in hell.mp3|:,1345760219,1345760545,1345795930,1345798229
  /mnt/sdb1/musikk/mp3/Twisted sister - Destroyer.mp3|    Twisted sister - Destroyer.mp3|:,1345763352,1346008426
  /mnt/sdb1/musikk/flac/Candlemass - At the gallows end.flac|    Candlemass - At the gallows end.flac|:,1345761248,1345796972,1346008710
  /mnt/sdb1/musikk/mp3/Saint deamon - My heart.mp3|    Saint deamon - My heart.mp3|:,1345761483,1345797207,1346008944
  /mnt/sdb1/musikk/mp3/Ræva Rockers - Depp (LP).mp3|    Ræva Rockers - Depp (LP).mp3|:,1345795310,1345797401,1346009139
To get it work I made a workaround and used awk separator |: instead of simply |
... I couldn't get the awk code to read column 3 instead of column 2 (as in the example in the main post.)
Some #comments would be great :D

The function in pMusic:

Code: Select all

-index_rating_buildlist)
	TIMESTAMP="$2"
	cat "$3" | 
	awk -F, -v TS=$TIMESTAMP '{
		  split($1, a, "|:"); $1=a[2] # fix first entry
		  max=NF; min=1;
		  
		  while (max-min > 1) {
			 i=int( (max+min)/2 )
			 if (TS >= $i) min = i
			 else max = i
		  }
		  print a[1], NF-min 
	}' 
	;;
Also, what we want is to define a alternative max value to find timestamps in a specific period. - Like february 2012. Let's say that is $4.


Anyone???
Sigmund

Awk: Count numbers in list

Posted: Mon 27 Aug 2012, 10:42
by L18L
@zigbert - faster than
read LINE; do ..... done < database.txt
is your
cat /root/database.txt | awk -F, -v TS=$FIND '...'

But even faster (10%) is:
awk -F, -v TS=$FIND ' ...' database.txt

Awk: Count numbers in list

Posted: Mon 27 Aug 2012, 12:05
by L18L
zigbert wrote:...My NEW index file looks right now like this:

Code: Select all

  /mnt/sdb1/musikk/mp3/Judas priest - Monsters of rock.mp3|    Judas priest - Monsters of rock.mp3|:,1345762237
To get it work I made a workaround and used awk separator |: instead of simply |
... I couldn't get the awk code to read column 3 instead of column 2 (as in the example in the main post.)
Some #comments would be great :D
Why 3 columns?
I have successfully tried to print file and basename from 2 column database:
changing just
print a[1], NF-min
to
num=split (a[1], b , "/")
print a[1], b[num], NF-min

Learning awk:
split(s, a [, r]) Splits the string s into the array a on the
regular expression r, and returns the number of
fields. If r is omitted, FS is used instead.
The array a is cleared first. Splitting behaves
identically to field splitting.

Go back to 2 column database :?:

Awk: Count numbers in list

Posted: Mon 27 Aug 2012, 12:52
by L18L
zigbert wrote:Also, what we want is to define a alternative max value to find timestamps in a specific period. - Like february 2012. Let's say that is $4.
Where is $1 ?

Here is a solution using $1 and $2 timestamp from and -to
#!/bin/ash
#additional output filename only=basename
#TIMESTAMP from to
#
LANG=C
#
FIND_FROM=$1 #timestamp
FIND_TO=$2

awk -F, -v TS1=$FIND_FROM -v TS2=$FIND_TO '{
split($1, a, "|"); $1=a[2] # fix first entry

max=NF; min=0;
while (max-min > 1) {
i=int( (max+min)/2 )
if (TS1 >= $i) min = i
else max = i
}
from=NF-min

max=NF; min=0;
while (max-min > 1) {
i=int( (max+min)/2 )
if (TS2 >= $i) min = i
else max = i
}
to=NF-min

num=split (a[1], b , "/")
print a[1], b[num], from-to

}' database.txt

Re: Awk: Count numbers in list

Posted: Mon 27 Aug 2012, 20:30
by zigbert
L18L wrote:But even faster (10%) is:
awk -F, -v TS=$FIND ' ...' database.txt
Done :D

Posted: Mon 27 Aug 2012, 20:32
by zigbert
L18L wrote:Why 3 columns?
Column 2 is not always basename.
Preferable it collects information from meta tags or m3u-file.

Re: Awk: Count numbers in list

Posted: Mon 27 Aug 2012, 21:01
by zigbert
L18L wrote:
zigbert wrote:Also, what we want is to define a alternative max value to find timestamps in a specific period. - Like february 2012. Let's say that is $4.
Where is $1 ? $1 is for pMusic internal redirection

Here is a solution using $1 and $2 timestamp from and -to
#!/bin/ash
#additional output filename only=basename
#TIMESTAMP from to
#
LANG=C
#
FIND_FROM=$1 #timestamp
FIND_TO=$2

awk -F, -v TS1=$FIND_FROM -v TS2=$FIND_TO '{
split($1, a, "|"); $1=a[2] # fix first entry

max=NF; min=0;
while (max-min > 1) {
i=int( (max+min)/2 )
if (TS1 >= $i) min = i
else max = i
}
from=NF-min

max=NF; min=0;
while (max-min > 1) {
i=int( (max+min)/2 )
if (TS2 >= $i) min = i
else max = i
}
to=NF-min
variable 'to' is never 0, even if zero hits

num=split (a[1], b , "/")
print a[1], b[num], from-to

}' database.txt
Else, I find this excellent!!!!!

Re: Awk: Count numbers in list

Posted: Tue 28 Aug 2012, 12:50
by L18L
zigbert wrote:Else, I find this excellent!!!!!
And our thanks go to: jamesbond :!:
All times are UTC
Page 1 of 2

Powered by phpBB® Forum Software © phpBB Limited