NetNews Usenet Archive 1992 #18/NN_1992_18.iso/spool/comp/unix/question/10026

Text File | 1992-08-12 | 5.3 KB | 214 lines

Newsgroups: comp.unix.questions
Path: sparky!uunet!zaphod.mps.ohio-state.edu!news.acns.nwu.edu!casbah.acns.nwu.edu!navarra
From: [e-mail] (John Navarra)
Subject: Associative arrays in [gn]awk --whynot! (long but interesting)
Message-ID: <[e-mail]>
Sender: [e-mail] (Usenet on news.acns)
Organization: Northwestern University, Evanston Illinois.
Date: Thu, 13 Aug 1992 06:33:48 GMT
Lines: 203

Well, after my previous posting on [gn]awk associative arrays I decided
to spend a little time investigating them. Here is what I came  up with.
I ran the following little awk script on a few files:
{
array[NR] = $1
}
END {
      for ( item in array)
          printf("%s\n", array[item] )
}


I ran it on files with 10,11,12,15 and 21 records where each record was
separated by a newline. Here are the results from the files with just
numbers in them. Basically, you can replace the numbers by anything
because no sorting is done by any of the programs before the arrays
are printed. We see some interesting behavior from both awk and nawk.
Gawk works as expected -- printing the array out as it was read in.
However, gawk did not work this way when more than one field was present
in a file. (more on that below)

awk   nawk    gawk
=====================10 file===============
2    2    1
3    3    2
4    4    3
5    5    4
6    6    5
7    7    6
8    8    7
9    9    8
10    10    9
1    1    10
=====================11 file===============
2    2    1
3    3    2
4    4    3
5    5    4
6    6    5
7    7    6
8    8    7
9    9    8
10    10    9
11    11    10
1    1    11
=====================12 file===============
2    2    1
3    3    2
4    4    3
5    5    4
6    6    5
7    7    6
8    8    7
9    9    8
10    10    9
11    11    10
12    12    11
1    1    12
=====================15 file===============
13    2    1
2    3    2
14    4    3
3    5    4
15    6    5
4    7    6
5    8    7
6    9    8
7    10    9
8    11    10
9    12    11
10    13    12
11    14    13
12    15    14
1    1    15
=====================21 file===============
13    2    1
2    3    2
14    4    3
3    5    4
15    6    5
4    7    6
16    8    7
5    9    8
17    10    9
6    11    10
18    12    11
7    13    12
19    14    13
8    15    14
9    16    15
10    17    16
20    18    17
11    19    18
21    20    19
12    21    20
1    1    21

======================
Here is a program that makes an associative array between the first and
fifth fields of a /etc/passwd-like file.

awk '
BEGIN { FS=":"}

{
#make asscociative array from first and fifth fields
fullname[$1]=$5
}
END {
   for ( username in fullname)
           printf("%-10s %s \n", username,  fullname[username]);
}' short.passwd


Here is the passwd-like file:
1root:##root:0:1:Operator:/:/bin/csh
2news:##news:6:6:Mr. News:/usr/local/lib/news:/bin/sh
3games:##games:11:10:Mr. Games:/usr/games:/bin/false
4ftp:##ftp:12:10:Mr. ftp:/usr/spool/ftp:/bin/false
5usenet:##usenet:20:12:Mr. Usenet:/var/spool/news:/bin/true
6server:*:21:10:Mr. Listserv:/usr/server:/bin/tcsh
7accman:##accman:284:30:Account
Manager:/usr/local/lib/cas:/usr/local/bin/cas
8nims:##nims:336:10:Christopher Nims:/home/u1/nims:/bin/tcsh
9matt:##matt:337:10:Matthew Larson:/home/u1/matt:/bin/tcsh
10mccoy:##mccoy:339:10:James McCoy:/home/u1/mccoy:/bin/tcsh
11pib:##pib:340:20:Philip Burns:/home/u1/pib:/bin/tcsh
12jln:##jln:341:20:John Norstad:/home/u1/jln:/bin/csh
13len:##len:343:20:Leonard Evens:/home/u1/len:/bin/csh
14dave:##dave:345:20:Dave Wenger:/home/u2/dave:/bin/tcsh
15navarra:##navarra:453:20:John Navarra:/home/u3/navarra:/bin/bash

Notice that I numbered the entries so you can see how out of order
they are with awk, nawk, and gawk versions:

awk:                results with just 15 numbers (from above)
1root      Operator        13
11pib      Philip Burns        2
5usenet    Mr. Usenet          14
7accman    Account Manager      3
14dave     Dave Wenger        15
6server    Mr. Listserv        4
13len      Leonard Evens    5
12jln      John Norstad        6
3games     Mr. Games        7
4ftp       Mr. ftp        8
10mccoy    James McCoy        9
9matt      Matthew Larson    10
8nims      Christopher Nims    11
2news      Mr. News        12
15navarra  John Navarra     1

nawk:
12jln      John Norstad        2
7accman    Account Manager    3
4ftp       Mr. ftp        4
10mccoy    James McCoy        5
14dave     Dave Wenger        6
11pib      Philip Burns        7
5usenet    Mr. Usenet        8
2news      Mr. News        9
6server    Mr. Listserv        10
1root      Operator        11
8nims      Christopher Nims    12
3games     Mr. Games        13
15navarra  John Navarra        14
9matt      Matthew Larson    15
13len      Leonard Evens    1

gawk:
10mccoy    James McCoy        1
6server    Mr. Listserv        2
5usenet    Mr. Usenet        3
15navarra  John Navarra        4
4ftp       Mr. ftp        5
12jln      John Norstad        6
9matt      Matthew Larson    7
11pib      Philip Burns        8
13len      Leonard Evens    9
14dave     Dave Wenger        10
3games     Mr. Games        11
8nims      Christopher Nims    12
2news      Mr. News        13
7accman    Account Manager    14
1root      Operator        15

    This is truly weird. I got the same results each time I ran
the program on the file for each of awk, nawk, and gawk. Each is completely
different from the other program and each is completely different from
its own result of 15 numbers from the test above.

I ran these test on a Sun running 4.1.2. If you want to try running the
above programs on your machine and send me the results, I would be
glad to look at them. Basically, associative arrays in gn[awk] are
unreliable (especially for a multi-field file). Gawk seems to be the
best at printing out arrays in the order which they were loaded in but
it does not do it correctly in all cases.

-tms
--
From the Lab of the MaD ScIenTiST:

[e-mail]