Prerequisites¶

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

  http://search.cpan.org/

DBI

A recent version of the DBI (Database independent interface for Perl). See below why.

DBD::File

This is the base class for DBD::CSV, and it is part of the DBI distribution. As DBD::CSV requires a matching version of DBD::File which is (partly) developed by the same team that maintains DBD::CSV. See META.json or Makefile.PL for the minimum versions.

SQL::Statement

A simple SQL engine. This module defines all of the SQL syntax for DBD::CSV, new SQL support is added with each release so you should look for updates to SQL::Statement regularly.

It is possible to run "DBD::CSV" without this module if you define the environment variable $DBI_SQL_NANO to 1. This will reduce the SQL support a lot though. See DBI::SQL::Nano for more details. Note that the test suite does only test in this mode in the development environment.

Text::CSV_XS

This module is used to read and write rows in a CSV file.

Installation¶

Installing this module (and the prerequisites from above) is quite simple. The simplest way is to install the bundle:

    $ cpan Bundle::DBD::CSV

Alternatively, you can name them all

    $ cpan Text::CSV_XS DBI DBD::CSV

or even trust "cpan" to resolve all dependencies for you:

    $ cpan DBD::CSV

If you cannot, for whatever reason, use cpan, fetch all modules from CPAN, and build with a sequence like:

    gzip -d < DBD-CSV-0.40.tgz | tar xf -

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

    cd DBD-CSV-0.40
    perl Makefile.PL
    make test

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

    make install UNINST=1

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.

Supported SQL Syntax¶

All SQL processing for DBD::CSV is done by SQL::Statement. See SQL::Statement for more specific information about its feature set. Features include joins, aliases, built-in and user-defined functions, and more. See SQL::Statement::Syntax for a description of the SQL syntax supported in DBD::CSV.

Table- and column-names are case insensitive unless quoted. Column names will be sanitized unless "raw_header" is true.

Creating a database handle (connect)¶

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 => "/home/user/folder",
        });

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

    $dbh = DBI->connect ("dbi:CSV:");
    $dbh = DBI->connect ("dbi:CSV:", undef, undef, { f_dir => "." });
    $dbh = DBI->connect ("dbi:CSV:f_dir=.");

We were told, that VMS might - for whatever reason - require:

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

The preferred way of passing the arguments is by driver attributes:

    # specify most possible flags via driver flags
    $dbh = DBI->connect ("dbi:CSV:", undef, undef, {
        f_schema         => undef,
        f_dir            => "data",
        f_dir_search     => [],
        f_ext            => ".csv/r",
        f_lock           => 2,
        f_encoding       => "utf8",
        csv_eol          => "\r\n",
        csv_sep_char     => ",",
        csv_quote_char   => '"',
        csv_escape_char  => '"',
        csv_class        => "Text::CSV_XS",
        csv_null         => 1,
        csv_bom          => 0,
        csv_tables       => {
            syspwd => {
                sep_char    => ":",
                quote_char  => undef,
                escape_char => undef,
                file        => "/etc/passwd",
                col_names   => [qw( login password
                                    uid gid realname
                                    directory shell )],
                },
            },
        RaiseError       => 1,
        PrintError       => 1,
        FetchHashKeyName => "NAME_lc",
        }) or die $DBI::errstr;

but you may set these attributes in the DSN as well, separated by semicolons. Pay attention to the semi-colon for "csv_sep_char" (as seen in many CSV exports from MS Excel) is being escaped in below example, as is would otherwise be seen as attribute separator:

    $dbh = DBI->connect (
        "dbi:CSV:f_dir=$ENV{HOME}/csvdb;f_ext=.csv;f_lock=2;" .
        "f_encoding=utf8;csv_eol=\n;csv_sep_char=\\;;" .
        "csv_quote_char=\";csv_escape_char=\\;csv_class=Text::CSV_XS;" .
        "csv_null=1") or die $DBI::errstr;

Using attributes in the DSN is easier to use when the DSN is derived from an outside source (environment variable, database entry, or configure file), whereas specifying entries in the attribute hash is easier to read and to maintain.

The default value for "csv_binary" is 1 (True).

The default value for "csv_auto_diag" is <1>. Note that this might cause trouble on perl versions older than 5.8.9, so up to and including perl version 5.8.8 it might be required to use ";csv_auto_diag=0" inside the "DSN" or "csv_auto_diag =" 0> inside the attributes.

Creating and dropping tables¶

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

    $dbh->do ("CREATE TABLE $table (id INTEGER, name CHAR (64))");
    $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 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.

Inserting, fetching and modifying data¶

The following examples insert some data in a table and fetch it back: First, an example where the column data is concatenated in the SQL 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 does not 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 quote column data passed as parameters. This version is particularly well designed for loops. Whenever performance is an issue, I recommend using this method.

You might wonder about the "undef". Don't wonder, just take it as it is. :-) It's an attribute argument that I have never used and will be passed to the prepare method as the 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 $sth = $dbh->prepare (qq;
        SELECT * FROM $table WHERE id > 1 ORDER BY id;
        ;);
    $sth->execute;
    my ($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 $sth = $dbh->prepare ("SELECT * FROM $table WHERE id = ?");
    $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 for details on these methods. See SQL::Statement 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");

Error handling¶

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

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

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

Metadata¶

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 "$sth->execute"

NUM_OF_PARAMS

Valid after "$sth->prepare"

NAME

NAME_lc

NAME_uc

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

NULLABLE

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

These attributes and methods are not supported:

    bind_param_inout
    CursorName
    LongReadLen
    LongTruncOk

DBD::File attributes¶

f_dir

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

f_dir_search

This attribute optionally defines a list of extra directories to search when opening existing tables. It should be an anonymous list or an array reference listing all folders where tables could be found.

    my $dbh = DBI->connect ("dbi:CSV:", "", "", {
        f_dir        => "data",
        f_dir_search => [ "ref/data", "ref/old" ],
        f_ext        => ".csv/r",
        }) or die $DBI::errstr;
    

f_ext

This attribute is used for setting the file extension.

f_schema

This attribute allows you to set the database schema name. The default is to use the owner of "f_dir". "undef" is allowed, but not in the DSN part.

    my $dbh = DBI->connect ("dbi:CSV:", "", "", {
        f_schema => undef,
        f_dir    => "data",
        f_ext    => ".csv/r",
        }) or die $DBI::errstr;
    

f_encoding

This attribute allows you to set the encoding of the data. With CSV, it is not possible to set (and remember) the encoding on a column basis, but DBD::File now allows the encoding to be set on the underlying file. If this attribute is not set, or undef is passed, the file will be seen as binary.

f_lock

With this attribute you can specify a locking mode to be used (if locking is supported at all) for opening tables. By default, tables are opened with a shared lock for reading, and with an exclusive lock for writing. The supported modes are:

But see "KNOWN BUGS" in DBD::File.

DBD::CSV specific attributes¶

csv_class

The attribute csv_class controls the CSV parsing engine. This defaults to "Text::CSV_XS", but "Text::CSV" can be used in some cases, too. Please be aware that "Text::CSV" does not care about any edge case as "Text::CSV_XS" does and that "Text::CSV" is probably about 100 times slower than "Text::CSV_XS".

In order to use the specified class other than "Text::CSV_XS", it needs to be loaded before use. "DBD::CSV" does not "require"/"use" the specified class itself.

Text::CSV_XS specific attributes¶

csv_eol

csv_sep_char

csv_quote_char

csv_escape_char

csv_csv

The attributes csv_eol, csv_sep_char, csv_quote_char and csv_escape_char are corresponding to the respective attributes of the csv_class (usually Text::CSV_CS) object. You may 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 csv_eol attribute defines the end-of-line pattern, which is better known as a record separator pattern since it separates records. The default is windows-style end-of-lines "\015\012" for output (writing) and unset for input (reading), so if on unix you may want to set this to newline ("\n") like this:

  $dbh->{csv_eol} = "\n";
    

It is also possible to use multi-character patterns as record separators. For example this file uses newlines as field separators (sep_char) and the pattern "\n__ENDREC__\n" as the record separators (eol):

  name
  city
  __ENDREC__
  joe
  seattle
  __ENDREC__
  sue
  portland
  __ENDREC__
    

To handle this file, you'd do this:

  $dbh->{eol}      = "\n__ENDREC__\n" ,
  $dbh->{sep_char} = "\n"
    

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_null

With this option set, all new statement handles will set "always_quote" and "blank_is_undef" in the CSV parser and writer, so it knows how to distinguish between the empty string and "undef" or "NULL". You cannot reset it with a false value. You can pass it to connect, or set it later:

  $dbh = DBI->connect ("dbi:CSV:", "", "", { csv_null => 1 });
  $dbh->{csv_null} = 1;
    

csv_bom

With this option set, the CSV parser will try to detect BOM (Byte Order Mark) in the header line. This requires Text::CSV_XS version 1.22 or higher.

  $dbh = DBI->connect ("dbi:CSV:", "", "", { csv_bom => 1 });
  $dbh->{csv_bom} = 1;
    

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:

  csv_tables => {
      syspwd => {                   # Table name
          csv_sep_char => ":",      # Text::CSV_XS
          quote_char   => undef,    # Text::CSV_XS
          escape_char  => undef,    # Text::CSV_XS
          f_dir        => "/etc",   # DBD::File
          f_file       => "passwd", # DBD::File
          col_names    =>           # DBD::File
            [qw( login password uid gid realname directory shell )],
          },
      },

csv_*

All other attributes that start with "csv_" and are not described above will be passed to "Text::CSV_XS" (without the "csv_" prefix). These extra options are only likely to be useful for reading (select) handles. Examples:

  $dbh->{csv_allow_whitespace}    = 1;
  $dbh->{csv_allow_loose_quotes}  = 1;
  $dbh->{csv_allow_loose_escapes} = 1;
    

See the "Text::CSV_XS" documentation for the full list and the documentation.

Driver specific attributes¶

f_file

The name of the file used for the table; 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 basis.

col_names

skip_first_row

By default DBD::CSV assumes that column names are stored in the first row of the CSV file and sanitizes them (see "raw_header" below). 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 "col0", "col1", ...

Note that column names that match reserved SQL words will cause unwanted and sometimes confusing errors. If your CSV has headers that match reserved words, you will require these two attributes.

If "test.csv" looks like

 select,from
 1,2
    

the select query would result in "select select, from from test;", which obviously is illegal SQL.

raw_header

Due to the SQL standard, field names cannot contain special characters like a dot (".") or a space (" ") unless the column names are quoted. Following the approach of mdb_tools, all these tokens are translated to an underscore ("_") when reading the first line of the CSV file, so all field names are 'sanitized'. If you do not want this to happen, set "raw_header" to a true value and the entries in the first line of the CSV data will be used verbatim for column headers and field names. DBD::CSV cannot guarantee that any part in the toolchain will work if field names have those characters, and the chances are high that the SQL statements will fail.

Currently, the sanitizing of headers is as simple as

  s/\W/_/g;
    

Note that headers (column names) might be folded in other parts of the code stack, specifically SQL::Statement, whose docs mention:

 Wildcards are expanded to lower cased identifiers. This might
 confuse some people, but it was easier to implement.
    

That means that in

 my $sth = $dbh->prepare ("select * from foo");
 $sth->execute;
 while (my $row = $sth->fetchrow_hashref) {
     say for keys %$row;
     }
    

all keys will show as all lower case, regardless of the original header.

It's strongly recommended to check the attributes supported by "Metadata" in DBD::File.

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

    use DBI;
    my $dbh = DBI->connect ("dbi:CSV:", undef, undef, {
        f_dir           => "/etc",
        csv_sep_char    => ":",
        csv_quote_char  => undef,
        csv_escape_char => undef,
        });
    $dbh->{csv_tables}{passwd} = {
        col_names => [qw( 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 override them on a per table basis:

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

Driver private methods¶

These methods are inherited from DBD::File:

data_sources

The "data_sources" method returns a list of sub-directories of the current directory in the form "dbi:CSV:directory=$dirname".

If you want to read the sub-directories 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 $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.