DBD::CSV - DBI driver for CSV files

    use DBI;
    $dbh = DBI->connect("DBI:CSV:f_dir=/home/joe/csvdb")
        or die "Cannot connect: " . $DBI::errstr;
    $sth = $dbh->prepare("CREATE TABLE a (id INTEGER, name CHAR(1))")
        or die "Cannot prepare: " . $dbh->errstr();
    $sth->execute() or die "Cannot execute: " . $sth->errstr();
    $sth->finish();
    $dbh->disconnect();

    # Read a CSV file with ";" as the separator, as exported by
    # MS Excel. Note we need to escape the ";", otherwise it
    # would be treated as an attribute separator.
    $dbh = DBI->connect(qq{DBI:CSV:csv_sep_char=\\;});
    $sth = $dbh->prepare("SELECT * FROM info");

    # Same example, this time reading "info.csv" as a table:
    $dbh = DBI->connect(qq{DBI:CSV:csv_sep_char=\\;});
    $dbh->{'csv_tables'}->{'info'} = { 'file' => 'info.csv'};
    $sth = $dbh->prepare("SELECT * FROM info");

The DBD::CSV module is yet another driver for the DBI (Database independent interface for Perl). This one is based on the SQL engine SQL::Statement and the abstract DBI driver DBD::File and implements access to so-called CSV files (Comma separated values). Such files are mostly used for exporting MS Access and MS Excel data.

See DBI(3) for details on DBI, SQL::Statement(3) for details on SQL::Statement and DBD::File(3) for details on the base class DBD::File.

The only system dependent feature that DBD::File uses, is the CWflock() function. Thus the module should run (in theory) on any system with a working CWflock(), in particular on all Unix machines and on Windows NT. Under Windows 95 and MacOS the use of CWflock() is disabled, thus the module should still be usable,

Unlike other DBI drivers, you don't need an external SQL engine or a running server. All you need are the following Perl modules, available from any CPAN mirror, for example

  ftp://ftp.funet.fi/pub/languages/perl/CPAN/modules/by-module

DBI

the DBI (Database independent interface for Perl), version 1.00 or a later release

SQL::Statement

a simple SQL engine

Text::CSV_XS

this module is used for writing rows to or reading rows from CSV files.

Installing this module (and the prerequisites from above) is quite simple. You just fetch the archive, extract it with

    gzip -cd DBD-CSV-0.1000.tar.gz | tar xf -

(this is for Unix users, Windows users would prefer WinZip or something similar) and then enter the following:

    cd DBD-CSV-0.1000
    perl Makefile.PL
    make
    make test

If any tests fail, let me know. Otherwise go on with

    make install

Note that you almost definitely need root or administrator permissions. If you don't have them, read the ExtUtils::MakeMaker man page for details on installing in your own directories. ExtUtils::MakeMaker.

  The level of SQL support available depends on the version of
  SQL::Statement installed.  Any version will support *basic*
  CREATE, INSERT, DELETE, UPDATE, and SELECT statements.  Only
  versions of SQL::Statement 1.0 and above support additional
  features such as table joins, string functions, etc.  See the
  documentation of the latest version of SQL::Statement for details.

Creating a database handle usually implies connecting to a database server. Thus this command reads

    use DBI;
    my $dbh = DBI->connect("DBI:CSV:f_dir=$dir");

The directory tells the driver where it should create or open tables (a.k.a. files). It defaults to the current directory, thus the following are equivalent:

    $dbh = DBI->connect("DBI:CSV:");
    $dbh = DBI->connect("DBI:CSV:f_dir=.");

(I was told, that VMS requires

    $dbh = DBI->connect("DBI:CSV:f_dir=");

for whatever reasons.)

You may set other attributes in the DSN string, separated by semicolons.

You can create and drop tables with commands like the following:

    $dbh->do("CREATE TABLE $table (id INTEGER, name CHAR(6))");
    $dbh->do("DROP TABLE $table");

Note that currently only the column names will be stored and no other data. Thus all other information including column type (INTEGER or CHAR(x), for example), column attributes (NOT NULL, PRIMARY KEY, ...) will silently be discarded. This may change in a later release.

A drop just removes the file without any warning.

See DBI(3) for more details.

Table names cannot be arbitrary, due to restrictions of the SQL syntax. I recommend that table names are valid SQL identifiers: The first character is alphabetic, followed by an arbitrary number of alphanumeric characters. If you want to use other files, the file names must start with '/', './' or '../' and they must not contain white space.

The following examples insert some data in a table and fetch it back: First all data in the string:

    $dbh->do("INSERT INTO $table VALUES (1, "
             . $dbh->quote("foobar") . ")");

Note the use of the quote method for escaping the word 'foobar'. Any string must be escaped, even if it doesn't contain binary data.

Next an example using parameters:

    $dbh->do("INSERT INTO $table VALUES (?, ?)", undef,
             2, "It's a string!");

Note that you don't need to use the quote method here, this is done automatically for you. This version is particularly well designed for loops. Whenever performance is an issue, I recommend using this method.

You might wonder about the CWundef. Don't wonder, just take it as it is. :-) It's an attribute argument that I have never ever used and will be parsed to the prepare method as a second argument.

To retrieve data, you can use the following:

    my($query) = "SELECT * FROM $table WHERE id > 1 ORDER BY id";
    my($sth) = $dbh->prepare($query);
    $sth->execute();
    while (my $row = $sth->fetchrow_hashref) {
        print("Found result row: id = ", $row->{'id'},
              ", name = ", $row->{'name'});
    }
    $sth->finish();

Again, column binding works: The same example again.

    my($query) = "SELECT * FROM $table WHERE id > 1 ORDER BY id";
    my($sth) = $dbh->prepare($query);
    $sth->execute();
    my($id, $name);
    $sth->bind_columns(undef, \$id, \$name);
    while ($sth->fetch) {
        print("Found result row: id = $id, name = $name\n");
    }
    $sth->finish();

Of course you can even use input parameters. Here's the same example for the third time:

    my($query) = "SELECT * FROM $table WHERE id = ?";
    my($sth) = $dbh->prepare($query);
    $sth->bind_columns(undef, \$id, \$name);
    for (my($i) = 1;  $i <= 2;   $i++) {
        $sth->execute($id);
        if ($sth->fetch) {
            print("Found result row: id = $id, name = $name\n");
        }
        $sth->finish();
    }

See DBI(3) for details on these methods. See SQL::Statement(3) for details on the WHERE clause.

Data rows are modified with the UPDATE statement:

    $dbh->do("UPDATE $table SET id = 3 WHERE id = 1");

Likewise you use the DELETE statement for removing rows:

    $dbh->do("DELETE FROM $table WHERE id > 1");

In the above examples we have never cared about return codes. Of course, this cannot be recommended. Instead we should have written (for example):

    my($query) = "SELECT * FROM $table WHERE id = ?";
    my($sth) = $dbh->prepare($query)
        or die "prepare: " . $dbh->errstr();
    $sth->bind_columns(undef, \$id, \$name)
        or die "bind_columns: " . $dbh->errstr();
    for (my($i) = 1;  $i <= 2;   $i++) {
        $sth->execute($id)
            or die "execute: " . $dbh->errstr();
        if ($sth->fetch) {
            print("Found result row: id = $id, name = $name\n");
        }
    }
    $sth->finish($id)
        or die "finish: " . $dbh->errstr();

Obviously this is tedious. Fortunately we have DBI's RaiseError attribute:

    $dbh->{'RaiseError'} = 1;
    $@ = '';
    eval {
        my($query) = "SELECT * FROM $table WHERE id = ?";
        my($sth) = $dbh->prepare($query);
        $sth->bind_columns(undef, \$id, \$name);
        for (my($i) = 1;  $i <= 2;   $i++) {
            $sth->execute($id);
            if ($sth->fetch) {
                print("Found result row: id = $id, name = $name\n");
            }
        }
        $sth->finish($id);
    };
    if ($@) { die "SQL database error: $@"; }

This is not only shorter, it even works when using DBI methods within subroutines.

The following attributes are handled by DBI itself and not by DBD::File, thus they all work as expected:

    Active
    ActiveKids
    CachedKids
    CompatMode             (Not used)
    InactiveDestroy
    Kids
    PrintError
    RaiseError
    Warn                   (Not used)

The following DBI attributes are handled by DBD::File:

AutoCommit

Always on

ChopBlanks

Works

NUM_OF_FIELDS

Valid after CW$sth->execute

NUM_OF_PARAMS

Valid after CW$sth->prepare

NAME

Valid after CW$sth->execute; undef for Non-Select statements.

NULLABLE

Not really working. Always returns an array ref of one's, as DBD::CSV doesn't verify input data. Valid after CW$sth->execute; undef for non-Select statements.

These attributes and methods are not supported:

    bind_param_inout
    CursorName
    LongReadLen
    LongTruncOk

In addition to the DBI attributes, you can use the following dbh attributes:

f_dir

This attribute is used for setting the directory where CSV files are opened. Usually you set it in the dbh, it defaults to the current directory (.). However, it is overwritable in the statement handles.

csv_eol

csv_sep_char

csv_quote_char

csv_escape_char

csv_class

csv_csv

The attributes csv_eol, csv_sep_char, csv_quote_char and csv_escape_char are corresponding to the respective attributes of the Text::CSV_XS object. You want to set these attributes if you have unusual CSV files like /etc/passwd or MS Excel generated CSV files with a semicolon as separator. Defaults are \015\012, ';', '' and '', respectively. The attributes are used to create an instance of the class csv_class, by default Text::CSV_XS. Alternatively you may pass an instance as csv_csv, the latter takes precedence. Note that the binary attribute must be set to a true value in that case. Additionally you may overwrite these attributes on a per-table base in the csv_tables attribute.

csv_tables

This hash ref is used for storing table dependent metadata. For any table it contains an element with the table name as key and another hash ref with the following attributes:

file

The tables file name; defaults to

    "$dbh->{f_dir}/$table"

eol

sep_char

quote_char

escape_char

class

csv

These correspond to the attributes csv_eol, csv_sep_char, csv_quote_char, csv_escape_char, csv_class and csv_csv. The difference is that they work on a per-table base.

col_names

skip_first_row

By default DBD::CSV assumes that column names are stored in the first row of the CSV file. If this is not the case, you can supply an array ref of table names with the col_names attribute. In that case the attribute skip_first_row will be set to FALSE. If you supply an empty array ref, the driver will read the first row for you, count the number of columns and create column names like CWcol0, CWcol1, ...

Example: Suggest you want to use /etc/passwd as a CSV file. :-) There simplest way is:

    require DBI;
    my $dbh = DBI->connect("DBI:CSV:f_dir=/etc;csv_eol=\n;"
                           . "csv_sep_char=:;csv_quote_char=;"
                           . "csv_escape_char=");
    $dbh->{'csv_tables'}->{'passwd'} = {
        'col_names' => ["login", "password", "uid", "gid", "realname",
                        "directory", "shell"]
    };
    $sth = $dbh->prepare("SELECT * FROM passwd");

Another possibility where you leave all the defaults as they are and overwrite them on a per table base:

    require DBI;
    my $dbh = DBI->connect("DBI:CSV:");
    $dbh->{'csv_tables'}->{'passwd'} = {
        'eol' => "\n",
        'sep_char' => ":",
        'quote_char' => undef,
        'escape_char' => undef,
        'file' => '/etc/passwd',
        'col_names' => ["login", "password", "uid", "gid", "realname",
                        "directory", "shell"]
    };
    $sth = $dbh->prepare("SELECT * FROM passwd");

These methods are inherited from DBD::File:

data_sources

The CWdata_sources method returns a list of subdirectories of the current directory in the form DBI:CSV:directory=$dirname. If you want to read the subdirectories of another directory, use

    my($drh) = DBI->install_driver("CSV");
    my(@list) = $drh->data_sources('f_dir' => '/usr/local/csv_data' );

list_tables

This method returns a list of file names inside CW$dbh->{'directory'}. Example:

    my($dbh) = DBI->connect("DBI:CSV:directory=/usr/local/csv_data");
    my(@list) = $dbh->func('list_tables');

Note that the list includes all files contained in the directory, even those that have non-valid table names, from the view of SQL. See Creating and dropping tables above.

When inserting and fetching data, you will sometimes be surprised: DBD::CSV doesn't correctly handle data types, in particular NULLs. If you insert integers, it might happen, that fetch returns a string. Of course, a string containing the integer, so that's perhaps not a real problem. But the following will never work:

    $dbh->do("INSERT INTO $table (id, name) VALUES (?, ?)",
             undef, "foo bar");
    $sth = $dbh->prepare("SELECT * FROM $table WHERE id IS NULL");
    $sth->execute();
    my($id, $name);
    $sth->bind_columns(undef, \$id, \$name);
    while ($sth->fetch) {
        printf("Found result row: id = %s, name = %s\n",
              defined($id) ? $id : "NULL",
              defined($name) ? $name : "NULL");
    }
    $sth->finish();

The row we have just inserted, will never be returned! The reason is obvious, if you examine the CSV file: The corresponding row looks like

    "","foo bar"

In other words, not a NULL is stored, but an empty string. CSV files don't have a concept of NULL values. Surprisingly the above example works, if you insert a NULL value for the name! Again, you find the explanation by examining the CSV file:

    ""

In other words, DBD::CSV has emulated a NULL value by writing a row with less columns. Of course this works only if the rightmost column is NULL, the two rightmost columns are NULL, ..., but the leftmost column will never be NULL!

See Creating and dropping tables above for table name restrictions.

Extensions of DBD::CSV:

CSV file scanner

Write a simple CSV file scanner that reads a CSV file and attempts to guess sep_char, quote_char, escape_char and eol automatically.

These are merely restrictions of the DBD::File or SQL::Statement modules:

Table name mapping

Currently it is not possible to use files with names like CWnames.csv. Instead you have to use soft links or rename files. As an alternative one might use, for example a dbh attribute 'table_map'. It might be a hash ref, the keys being the table names and the values being the file names.

Column name mapping

Currently the module assumes that column names are stored in the first row. While this is fine in most cases, there should be a possibility of setting column names and column number from the programmer: For example MS Access doesn't export column names by default.

•

The module is using flock() internally. However, this function is not available on platforms. Using flock() is disabled on MacOS and Windows 95: There's no locking at all (perhaps not so important on these operating systems, as they are for single users anyways).

This module is currently maintained by

      Jeff Zucker
      <jeff@vpservices.com>

The original author is Jochen Wiedmann.

Copyright (C) 1998 by Jochen Wiedmann

All rights reserved.

You may distribute this module under the terms of either the GNU General Public License or the Artistic License, as specified in the Perl README file.

DBI(3), Text::CSV_XS(3), SQL::Statement(3)

For help on the use of DBD::CSV, see the DBI users mailing list:

  http://www.isc.org/dbi-lists.html

For general information on DBI see

  http://www.symbolstone.org/technology/perl/DBI