NetNews Usenet Archive 1992 #20/NN_1992_20.iso/spool/comp/lang/perl/5762

Internet Message Format | 1992-09-08 | 3.7 KB

Path: sparky!uunet!pipex!unipalm!uknet!doc.ic.ac.uk!doc.ic.ac.uk!not-for-mail
From: [e-mail] (Diomidis D Spinellis)
Newsgroups: comp.lang.perl
Subject: Re: Using perl for databases - experiences requested
Keywords: database implementation experience
Message-ID: <[e-mail]>
Date: 5 Sep 92 18:59:09 GMT
References: <[e-mail]>
Organization: Department of Computing, Imperial College, University of London, UK.
Lines: 83
NNTP-Posting-Host: swan.doc.ic.ac.uk

In article <[e-mail]> [e-mail] (Scott Meyer) writes:
>What sort of experience have people had using perl to implement databases?

I have used Perl to implement two databases.

The first database is a small system I use for my thesis to keep track
of various language characteristics.  I save the data using the same
approach as the one you describe in your article.  Using single
character field identification strings allows me to parse each record
into an assiociative array of strings using the following loop:

        @rec = split(/\n\%/);
        $rec[0] =~ s/^\%//;
        foreach $r (@rec) {
                ($key, $val) = ($r =~ m/(.) (.*)$/);
                $rec{$key} = $val;
        }

So for a record like:

    %N Perl
    %A Larry Wall
    %I Interpreted

$rec{N} will be Perl, $rec{A} will be 'Larry Wall', and $rec{I} will be
'Interpreted'.  Records are separated by single blank lines, so I can
simply use $/=''; $*=1 to parse the input.  You mention in your article
that you use filters to process complex queries.  I tried to write the
query processing in the database script, but found myself to prefer the
filter approach, which I am now using.  I feel uneasy about using
pipelines of Perl scripts (some complex queries clock up the process id
counter up by as much as 400 in an otherwise idle machine), but it
seems to work fine up to now.  Most query scripts use the dbgrep Perl
script which greps patterns in the database.  It is a very small script
whose engine consists of the following lines:

    $/=''; $*=1;
    $pattern = shift;
        eval "
        while (<>) {
                print if ($pattern);
        }
        ";

The initial Perl query processing code was based on queries in Perl syntax
of the type Field-letter = value.  These were changed into Perl by the
command:

    s?([$REC])\s*=\s*'([^']+)'?(((\$i) = m/%$1([^%]*)/) && \$i =~ m/$2/)?g;

and then put into an array using the command:

    $x = $_; eval "\@selected = grep($x, \@db);";

For example the query:

    select N='C' && C='Functional'

will generate the following expression to be passed through grep:

    ((($i) = m/%N([^%]*)/) && $i =~ m/C/) &&
    ((($i) = m/%C([^%]*)/) && $i =~ m/Functional/)

($i is first assigned the whole field needed, and then is matched
against its contents).  The approach worked without a problem, but I
found scripts much more convenient.  I wish Perl had the capability to
pipe through control structures like the Bourne Shell.

The second database is a medical information system used in a hospital
to track patient data.  Its user-interface is implemented in C.  The
data is stored in gdbm files.  Perl scripts are used to handle query
handling and report generation.  Queries are parsed from the `natural
language' (do not ask me for it, you don't want it: it parses commands
expressed in Greek ) syntax into a Perl expression, within a loop,
which is then eval'ed.  The records are variable sized, with a key
generated from the name together with a unique number.  Last time I
checked there were more than 800 records in the database, and Perl was
handling it without a problem.

Diomidis
--
Diomidis Spinellis    Internet: <[e-mail]>  UUCP: ...!uknet!icdoc!dds
Department of Computing, Imperial College, London SW7     #include "/dev/tty"