• NAME
• SYNOPSIS
• DESCRIPTION
  • How does DB_File interface to Berkeley DB?
  • Opening a Berkeley DB Database File
  • Default Parameters
  • In Memory Databases
• DB_HASH
  • A Simple Example.
• DB_BTREE
  • Changing the BTREE sort order
  • Handling duplicate keys
  • The get_dup method.
  • Matching Partial Keys
• DB_RECNO
  • The bval option
  • A Simple Example
  • Extra Methods
  • Another Example
• THE API INTERFACE
• HINTS AND TIPS
  • Locking Databases
  • Sharing databases with C applications
  • The untie gotcha
• COMMON QUESTIONS
  • Why is there Perl source in my database?
  • How do I store complex data structures with DB_File?
  • What does "Invalid Argument" mean?
  • What does "Bareword 'DB_File' not allowed" mean?
• HISTORY
• BUGS
• AVAILABILITY
• SEE ALSO
• AUTHOR

NAME

SYNOPSIS

 use DB_File ;
 
 [$X =] tie %hash,  'DB_File', [$filename, $flags, $mode, $DB_HASH] ;
 [$X =] tie %hash,  'DB_File', $filename, $flags, $mode, $DB_BTREE ;
 [$X =] tie @array, 'DB_File', $filename, $flags, $mode, $DB_RECNO ;

 $status = $X->del($key [, $flags]) ;
 $status = $X->put($key, $value [, $flags]) ;
 $status = $X->get($key, $value [, $flags]) ;
 $status = $X->seq($key, $value, $flags) ;
 $status = $X->sync([$flags]) ;
 $status = $X->fd ;

 # BTREE only
 $count = $X->get_dup($key) ;
 @list  = $X->get_dup($key) ;
 %list  = $X->get_dup($key, 1) ;

 # RECNO only
 $a = $X->length;
 $a = $X->pop ;
 $X->push(list);
 $a = $X->shift;
 $X->unshift(list);

 untie %hash ;
 untie @array ;

DESCRIPTION

Berkeley DB is a C library which provides a consistent interface to a number of database formats. DB_File provides an interface to all three of the database types currently supported by Berkeley DB.

The file types are:

DB_HASH

This database type allows arbitrary key/value pairs to be stored in data files. This is equivalent to the functionality provided by other hashing packages like DBM, NDBM, ODBM, GDBM, and SDBM. Remember though, the files created using DB_HASH are not compatible with any of the other packages mentioned.

A default hashing algorithm, which will be adequate for most applications, is built into Berkeley DB. If you do need to use your own hashing algorithm it is possible to write your own in Perl and have DB_File use it instead.

DB_BTREE

The btree format allows arbitrary key/value pairs to be stored in a sorted, balanced binary tree.

As with the DB_HASH format, it is possible to provide a user defined Perl routine to perform the comparison of keys. By default, though, the keys are stored in lexical order.

DB_RECNO

DB_RECNO allows both fixed-length and variable-length flat text files to be manipulated using the same key/value pair interface as in DB_HASH and DB_BTREE. In this case the key will consist of a record (line) number.

How does DB_File interface to Berkeley DB?

In addition to the tie interface, it is also possible to access most of the functions provided in the Berkeley DB API directly. See THE API INTERFACE.

Opening a Berkeley DB Database File

      DB*
      dbopen (const char * file, int flags, int mode, 
              DBTYPE type, const void * openinfo)

The parameter type is an enumeration which specifies which of the 3 interface methods (DB_HASH, DB_BTREE or DB_RECNO) is to be used. Depending on which of these is actually chosen, the final parameter, openinfo points to a data structure which allows tailoring of the specific interface method.

This interface is handled slightly differently in DB_File. Here is an equivalent call using DB_File:

        tie %array, 'DB_File', $filename, $flags, $mode, $DB_HASH ;

The filename, flags and mode parameters are the direct equivalent of their dbopen counterparts. The final parameter $DB_HASH performs the function of both the type and openinfo parameters in dbopen.

In the example above $DB_HASH is actually a pre-defined reference to a hash object. DB_File has three of these pre-defined references. Apart from $DB_HASH, there is also $DB_BTREE and $DB_RECNO.

The keys allowed in each of these pre-defined references is limited to the names used in the equivalent C structure. So, for example, the $DB_HASH reference will only allow keys called bsize, cachesize, ffactor, hash, lorder and nelem.

To change one of these elements, just assign to it like this:

	$DB_HASH->{'cachesize'} = 10000 ;

The three predefined variables $DB_HASH, $DB_BTREE and $DB_RECNO are usually adequate for most applications. If you do need to create extra instances of these objects, constructors are available for each file type.

Here are examples of the constructors and the valid options available for DB_HASH, DB_BTREE and DB_RECNO respectively.

     $a = new DB_File::HASHINFO ;
     $a->{'bsize'} ;
     $a->{'cachesize'} ;
     $a->{'ffactor'};
     $a->{'hash'} ;
     $a->{'lorder'} ;
     $a->{'nelem'} ;

     $b = new DB_File::BTREEINFO ;
     $b->{'flags'} ;
     $b->{'cachesize'} ;
     $b->{'maxkeypage'} ;
     $b->{'minkeypage'} ;
     $b->{'psize'} ;
     $b->{'compare'} ;
     $b->{'prefix'} ;
     $b->{'lorder'} ;

     $c = new DB_File::RECNOINFO ;
     $c->{'bval'} ;
     $c->{'cachesize'} ;
     $c->{'psize'} ;
     $c->{'flags'} ;
     $c->{'lorder'} ;
     $c->{'reclen'} ;
     $c->{'bfname'} ;

The values stored in the hashes above are mostly the direct equivalent of their C counterpart. Like their C counterparts, all are set to a default values - that means you don't have to set all of the values when you only want to change one. Here is an example:

     $a = new DB_File::HASHINFO ;
     $a->{'cachesize'} =  12345 ;
     tie %y, 'DB_File', "filename", $flags, 0777, $a ;

A few of the options need extra discussion here. When used, the C equivalent of the keys hash, compare and prefix store pointers to C functions. In DB_File these keys are used to store references to Perl subs. Below are templates for each of the subs:

    sub hash
    {
        my ($data) = @_ ;
        ...
        # return the hash value for $data
	return $hash ;
    }

    sub compare
    {
	my ($key, $key2) = @_ ;
        ...
        # return  0 if $key1 eq $key2
        #        -1 if $key1 lt $key2
        #         1 if $key1 gt $key2
        return (-1 , 0 or 1) ;
    }

    sub prefix
    {
	my ($key, $key2) = @_ ;
        ...
        # return number of bytes of $key2 which are 
        # necessary to determine that it is greater than $key1
        return $bytes ;
    }

See Changing the BTREE sort order for an example of using the compare template.

If you are using the DB_RECNO interface and you intend making use of bval, you should check out The bval option.

Default Parameters

    tie %A, "DB_File", "filename" ;

is equivalent to:

    tie %A, "DB_File", "filename", O_CREAT|O_RDWR, 0666, $DB_HASH ;

It is also possible to omit the filename parameter as well, so the call:

    tie %A, "DB_File" ;

is equivalent to:

    tie %A, "DB_File", undef, O_CREAT|O_RDWR, 0666, $DB_HASH ;

See In Memory Databases for a discussion on the use of undef in place of a filename.

In Memory Databases

DB_HASH

A Simple Example.

    use strict ;
    use DB_File ;
    use vars qw( %h $k $v ) ;

    tie %h, "DB_File", "fruit", O_RDWR|O_CREAT, 0640, $DB_HASH 
        or die "Cannot open file 'fruit': $!\n";

    # Add a few key/value pairs to the file
    $h{"apple"} = "red" ;
    $h{"orange"} = "orange" ;
    $h{"banana"} = "yellow" ;
    $h{"tomato"} = "red" ;

    # Check for existence of a key
    print "Banana Exists\n\n" if $h{"banana"} ;

    # Delete a key/value pair.
    delete $h{"apple"} ;

    # print the contents of the file
    while (($k, $v) = each %h)
      { print "$k -> $v\n" }

    untie %h ;

here is the output:

    Banana Exists
 
    orange -> orange
    tomato -> red
    banana -> yellow

Note that the like ordinary associative arrays, the order of the keys retrieved is in an apparently random order.

DB_BTREE

Changing the BTREE sort order

    use strict ;
    use DB_File ;

    my %h ;

    sub Compare
    {
        my ($key1, $key2) = @_ ;
        "\L$key1" cmp "\L$key2" ;
    }

    # specify the Perl sub that will do the comparison
    $DB_BTREE->{'compare'} = \&Compare ;

    tie %h, "DB_File", "tree", O_RDWR|O_CREAT, 0640, $DB_BTREE 
        or die "Cannot open file 'tree': $!\n" ;

    # Add a key/value pair to the file
    $h{'Wall'} = 'Larry' ;
    $h{'Smith'} = 'John' ;
    $h{'mouse'} = 'mickey' ;
    $h{'duck'}  = 'donald' ;

    # Delete
    delete $h{"duck"} ;

    # Cycle through the keys printing them in order.
    # Note it is not necessary to sort the keys as
    # the btree will have kept them in order automatically.
    foreach (keys %h)
      { print "$_\n" }

    untie %h ;

Here is the output from the code above.

    mouse
    Smith
    Wall

There are a few point to bear in mind if you want to change the ordering in a BTREE database:

1. The new compare function must be specified when you create the database.

2. You cannot change the ordering once the database has been created. Thus you must use the same compare function every time you access the database.

Handling duplicate keys

There are some difficulties in using the tied hash interface if you want to manipulate a BTREE database with duplicate keys. Consider this code:

    use strict ;
    use DB_File ;

    use vars qw($filename %h ) ;

    $filename = "tree" ;
    unlink $filename ;
 
    # Enable duplicate records
    $DB_BTREE->{'flags'} = R_DUP ;
 
    tie %h, "DB_File", $filename, O_RDWR|O_CREAT, 0640, $DB_BTREE 
	or die "Cannot open $filename: $!\n";
 
    # Add some key/value pairs to the file
    $h{'Wall'} = 'Larry' ;
    $h{'Wall'} = 'Brick' ; # Note the duplicate key
    $h{'Wall'} = 'Brick' ; # Note the duplicate key and value
    $h{'Smith'} = 'John' ;
    $h{'mouse'} = 'mickey' ;

    # iterate through the associative array
    # and print each key/value pair.
    foreach (keys %h)
      { print "$_  -> $h{$_}\n" }

    untie %h ;

Here is the output:

    Smith   -> John
    Wall    -> Larry
    Wall    -> Larry
    Wall    -> Larry
    mouse   -> mickey

As you can see 3 records have been successfully created with key Wall - the only thing is, when they are retrieved from the database they seem to have the same value, namely Larry. The problem is caused by the way that the associative array interface works. Basically, when the associative array interface is used to fetch the value associated with a given key, it will only ever retrieve the first value.

Although it may not be immediately obvious from the code above, the associative array interface can be used to write values with duplicate keys, but it cannot be used to read them back from the database.

The way to get around this problem is to use the Berkeley DB API method called seq. This method allows sequential access to key/value pairs. See THE API INTERFACE for details of both the seq method and the API in general.

Here is the script above rewritten using the seq API method.

    use strict ;
    use DB_File ;
 
    use vars qw($filename $x %h $status $key $value) ;

    $filename = "tree" ;
    unlink $filename ;
 
    # Enable duplicate records
    $DB_BTREE->{'flags'} = R_DUP ;
 
    $x = tie %h, "DB_File", $filename, O_RDWR|O_CREAT, 0640, $DB_BTREE 
	or die "Cannot open $filename: $!\n";
 
    # Add some key/value pairs to the file
    $h{'Wall'} = 'Larry' ;
    $h{'Wall'} = 'Brick' ; # Note the duplicate key
    $h{'Wall'} = 'Brick' ; # Note the duplicate key and value
    $h{'Smith'} = 'John' ;
    $h{'mouse'} = 'mickey' ;
 
    # iterate through the btree using seq
    # and print each key/value pair.
    $key = $value = 0 ;
    for ($status = $x->seq($key, $value, R_FIRST) ;
         $status == 0 ;
         $status = $x->seq($key, $value, R_NEXT) )
      {  print "$key -> $value\n" }
 
    undef $x ;
    untie %h ;

that prints:

    Smith   -> John
    Wall    -> Brick
    Wall    -> Brick
    Wall    -> Larry
    mouse   -> mickey

This time we have got all the key/value pairs, including the multiple values associated with the key Wall.

The get_dup method.

    $count = $x->get_dup($key) ;
    @list  = $x->get_dup($key) ;
    %list  = $x->get_dup($key, 1) ;

In a scalar context the method returns the number of values associated with the key, $key.

In list context, it returns all the values which match $key. Note that the values will be returned in an apparently random order.

In list context, if the second parameter is present and evaluates TRUE, the method returns an associative array. The keys of the associative array correspond to the the values that matched in the BTREE and the values of the array are a count of the number of times that particular value occurred in the BTREE.

So assuming the database created above, we can use get_dup like this:

    my $cnt  = $x->get_dup("Wall") ;
    print "Wall occurred $cnt times\n" ;

    my %hash = $x->get_dup("Wall", 1) ;
    print "Larry is there\n" if $hash{'Larry'} ;
    print "There are $hash{'Brick'} Brick Walls\n" ;

    my @list = $x->get_dup("Wall") ;
    print "Wall =>	[@list]\n" ;

    @list = $x->get_dup("Smith") ;
    print "Smith =>	[@list]\n" ;
 
    @list = $x->get_dup("Dog") ;
    print "Dog =>	[@list]\n" ;

and it will print:

    Wall occurred 3 times
    Larry is there
    There are 2 Brick Walls
    Wall =>	[Brick Brick Larry]
    Smith =>	[John]
    Dog =>	[]

Matching Partial Keys

    $x->seq($key, $value, R_CURSOR) ;

Here is the relevant quote from the dbopen man page where it defines the use of the R_CURSOR flag with seq:

    Note, for the DB_BTREE access method, the returned key is not
    necessarily an exact match for the specified key. The returned key
    is the smallest key greater than or equal to the specified key,
    permitting partial key matches and range searches.

In the example script below, the match sub uses this feature to find and print the first matching key/value pair given a partial key.

    use strict ;
    use DB_File ;
    use Fcntl ;

    use vars qw($filename $x %h $st $key $value) ;

    sub match
    {
        my $key = shift ;
        my $value = 0;
        my $orig_key = $key ;
        $x->seq($key, $value, R_CURSOR) ;
        print "$orig_key\t-> $key\t-> $value\n" ;
    }

    $filename = "tree" ;
    unlink $filename ;

    $x = tie %h, "DB_File", $filename, O_RDWR|O_CREAT, 0640, $DB_BTREE
        or die "Cannot open $filename: $!\n";
 
    # Add some key/value pairs to the file
    $h{'mouse'} = 'mickey' ;
    $h{'Wall'} = 'Larry' ;
    $h{'Walls'} = 'Brick' ; 
    $h{'Smith'} = 'John' ;
 

    $key = $value = 0 ;
    print "IN ORDER\n" ;
    for ($st = $x->seq($key, $value, R_FIRST) ;
	 $st == 0 ;
         $st = $x->seq($key, $value, R_NEXT) )
	
      {  print "$key -> $value\n" }
 
    print "\nPARTIAL MATCH\n" ;

    match "Wa" ;
    match "A" ;
    match "a" ;

    undef $x ;
    untie %h ;

Here is the output:

    IN ORDER
    Smith -> John
    Wall  -> Larry
    Walls -> Brick
    mouse -> mickey

    PARTIAL MATCH
    Wa -> Wall  -> Larry
    A  -> Smith -> John
    a  -> mouse -> mickey

DB_RECNO

In order to make RECNO more compatible with Perl the array offset for all RECNO arrays begins at 0 rather than 1 as in Berkeley DB.

As with normal Perl arrays, a RECNO array can be accessed using negative indexes. The index -1 refers to the last element of the array, -2 the second last, and so on. Attempting to access an element before the start of the array will raise a fatal run-time error.

The bval option

    The delimiting byte to be used to mark  the  end  of  a
    record for variable-length records, and the pad charac-
    ter for fixed-length records.  If no  value  is  speci-
    fied,  newlines  (``\n'')  are  used to mark the end of
    variable-length records and  fixed-length  records  are
    padded with spaces.

The second sentence is wrong. In actual fact bval will only default to "\n" when the openinfo parameter in dbopen is NULL. If a non-NULL openinfo parameter is used at all, the value that happens to be in bval will be used. That means you always have to specify bval when making use of any of the options in the openinfo parameter. This documentation error will be fixed in the next release of Berkeley DB.

That clarifies the situation with regards Berkeley DB itself. What about DB_File? Well, the behavior defined in the quote above is quite useful, so DB_File conforms it.

That means that you can specify other options (e.g. cachesize) and still have bval default to "\n" for variable length records, and space for fixed length records.

A Simple Example

    use strict ;
    use DB_File ;

    my @h ;
    tie @h, "DB_File", "text", O_RDWR|O_CREAT, 0640, $DB_RECNO 
        or die "Cannot open file 'text': $!\n" ;

    # Add a few key/value pairs to the file
    $h[0] = "orange" ;
    $h[1] = "blue" ;
    $h[2] = "yellow" ;

    # Check for existence of a key
    print "Element 1 Exists with value $h[1]\n" if $h[1] ;

    # use a negative index
    print "The last element is $h[-1]\n" ;
    print "The 2nd last element is $h[-2]\n" ;

    untie @h ;

Here is the output from the script:

    Element 1 Exists with value blue
    The last element is yellow
    The 2nd last element is blue

Extra Methods

Here are the methods:

$X-gtpush(list) ;

Pushes the elements of list to the end of the array.

$value = $X-gtpop ;

Removes and returns the last element of the array.

$X-gtshift

Removes and returns the first element of the array.

$X-gtunshift(list) ;

Pushes the elements of list to the start of the array.

$X-gtlength

Returns the number of elements in the array.

Another Example

    use strict ;
    use vars qw(@h $H $file $i) ;
    use DB_File ;
    use Fcntl ;
    
    $file = "text" ;

    unlink $file ;

    $H = tie @h, "DB_File", $file, O_RDWR|O_CREAT, 0640, $DB_RECNO 
        or die "Cannot open file $file: $!\n" ;
    
    # first create a text file to play with
    $h[0] = "zero" ;
    $h[1] = "one" ;
    $h[2] = "two" ;
    $h[3] = "three" ;
    $h[4] = "four" ;

    
    # Print the records in order.
    #
    # The length method is needed here because evaluating a tied
    # array in a scalar context does not return the number of
    # elements in the array.  

    print "\nORIGINAL\n" ;
    foreach $i (0 .. $H->length - 1) {
        print "$i: $h[$i]\n" ;
    }

    # use the push & pop methods
    $a = $H->pop ;
    $H->push("last") ;
    print "\nThe last record was [$a]\n" ;

    # and the shift & unshift methods
    $a = $H->shift ;
    $H->unshift("first") ;
    print "The first record was [$a]\n" ;

    # Use the API to add a new record after record 2.
    $i = 2 ;
    $H->put($i, "Newbie", R_IAFTER) ;

    # and a new record before record 1.
    $i = 1 ;
    $H->put($i, "New One", R_IBEFORE) ;

    # delete record 3
    $H->del(3) ;

    # now print the records in reverse order
    print "\nREVERSE\n" ;
    for ($i = $H->length - 1 ; $i >= 0 ; -- $i)
      { print "$i: $h[$i]\n" }

    # same again, but use the API functions instead
    print "\nREVERSE again\n" ;
    my ($s, $k, $v)  = (0, 0, 0) ;
    for ($s = $H->seq($k, $v, R_LAST) ; 
             $s == 0 ; 
             $s = $H->seq($k, $v, R_PREV))
      { print "$k: $v\n" }

    undef $H ;
    untie @h ;

and this is what it outputs:

    ORIGINAL
    0: zero
    1: one
    2: two
    3: three
    4: four

    The last record was [four]
    The first record was [zero]

    REVERSE
    5: last
    4: three
    3: Newbie
    2: one
    1: New One
    0: first

    REVERSE again
    5: last
    4: three
    3: Newbie
    2: one
    1: New One
    0: first

Notes:

1. Rather than iterating through the array, @h like this:

       foreach $i (@h)
   

   it is necessary to use either this:

       foreach $i (0 .. $H->length - 1) 
   

   or this:

       for ($a = $H->get($k, $v, R_FIRST) ;
            $a == 0 ;
            $a = $H->get($k, $v, R_NEXT) )
   

2. Notice that both times the put method was used the record index was specified using a variable, $i, rather than the literal value itself. This is because put will return the record number of the inserted line via that parameter.

THE API INTERFACE

To do this you need to store a copy of the object returned from the tie.

	$db = tie %hash, "DB_File", "filename" ;

Once you have done that, you can access the Berkeley DB API functions as DB_File methods directly like this:

	$db->put($key, $value, R_NOOVERWRITE) ;

Important: If you have saved a copy of the object returned from tie, the underlying database file will not be closed until both the tied variable is untied and all copies of the saved object are destroyed.

    use DB_File ;
    $db = tie %hash, "DB_File", "filename" 
        or die "Cannot tie filename: $!" ;
    ...
    undef $db ;
    untie %hash ;

See The untie Gotcha for more details.

All the functions defined in dbopen are available except for close and dbopen itself. The DB_File method interface to the supported functions have been implemented to mirror the way Berkeley DB works whenever possible. In particular note that:

• The methods return a status value. All return 0 on success. All return -1 to signify an error and set $! to the exact error code. The return code 1 generally (but not always) means that the key specified did not exist in the database.

  Other return codes are defined. See below and in the Berkeley DB documentation for details. The Berkeley DB documentation should be used as the definitive source.

• Whenever a Berkeley DB function returns data via one of its parameters, the equivalent DB_File method does exactly the same.

• If you are careful, it is possible to mix API calls with the tied hash/array interface in the same piece of code. Although only a few of the methods used to implement the tied interface currently make use of the cursor, you should always assume that the cursor has been changed any time the tied hash/array interface is used. As an example, this code will probably not do what you expect:

      $X = tie %x, 'DB_File', $filename, O_RDWR|O_CREAT, 0777, $DB_BTREE
          or die "Cannot tie $filename: $!" ;
  

      # Get the first key/value pair and set  the cursor
      $X->seq($key, $value, R_FIRST) ;
  

      # this line will modify the cursor
      $count = scalar keys %x ; 
  

      # Get the second key/value pair.
      # oops, it didn't, it got the last key/value pair!
      $X->seq($key, $value, R_NEXT) ;
  

  The code above can be rearranged to get around the problem, like this:

      $X = tie %x, 'DB_File', $filename, O_RDWR|O_CREAT, 0777, $DB_BTREE
          or die "Cannot tie $filename: $!" ;
  

      # this line will modify the cursor
      $count = scalar keys %x ; 
  

      # Get the first key/value pair and set  the cursor
      $X->seq($key, $value, R_FIRST) ;
  

      # Get the second key/value pair.
      # worked this time.
      $X->seq($key, $value, R_NEXT) ;
  

Below is a list of the methods available.

$status = $X-gtget($key, $value [, $flags]) ;

Given a key ($key) this method reads the value associated with it from the database. The value read from the database is returned in the $value parameter.

If the key does not exist the method returns 1.

No flags are currently defined for this method.

$status = $X-gtput($key, $value [, $flags]) ;

Stores the key/value pair in the database.

If you use either the R_IAFTER or R_IBEFORE flags, the $key parameter will have the record number of the inserted key/value pair set.

Valid flags are R_CURSOR, R_IAFTER, R_IBEFORE, R_NOOVERWRITE and R_SETCURSOR.

$status = $X-gtdel($key [, $flags]) ;

Removes all key/value pairs with key $key from the database.

A return code of 1 means that the requested key was not in the database.

R_CURSOR is the only valid flag at present.

$status = $X-gtfd ;

Returns the file descriptor for the underlying database.

See Locking Databases for an example of how to make use of the fd method to lock your database.

$status = $X-gtseq($key, $value, $flags) ;

This interface allows sequential retrieval from the database. See dbopen for full details.

Both the $key and $value parameters will be set to the key/value pair read from the database.

The flags parameter is mandatory. The valid flag values are R_CURSOR, R_FIRST, R_LAST, R_NEXT and R_PREV.

$status = $X-gtsync([$flags]) ;

Flushes any cached buffers to disk.

R_RECNOSYNC is the only valid flag at present.

HINTS AND TIPS

Locking Databases

    use DB_File;

    use strict;

    sub LOCK_SH { 1 }
    sub LOCK_EX { 2 }
    sub LOCK_NB { 4 }
    sub LOCK_UN { 8 }

    my($oldval, $fd, $db, %db, $value, $key);

    $key = shift || 'default';
    $value = shift || 'magic';

    $value .= " $$";

    $db = tie(%db, 'DB_File', '/tmp/foo.db', O_CREAT|O_RDWR, 0644) 
	    || die "dbcreat /tmp/foo.db $!";
    $fd = $db->fd;
    print "$$: db fd is $fd\n";
    open(DB_FH, "+<&=$fd") || die "dup $!";

    unless (flock (DB_FH, LOCK_SH | LOCK_NB)) {
	print "$$: CONTENTION; can't read during write update!
		    Waiting for read lock ($!) ....";
	unless (flock (DB_FH, LOCK_SH)) { die "flock: $!" }
    } 
    print "$$: Read lock granted\n";

    $oldval = $db{$key};
    print "$$: Old value was $oldval\n";
    flock(DB_FH, LOCK_UN);

    unless (flock (DB_FH, LOCK_EX | LOCK_NB)) {
	print "$$: CONTENTION; must have exclusive lock!
		    Waiting for write lock ($!) ....";
	unless (flock (DB_FH, LOCK_EX)) { die "flock: $!" }
    } 

    print "$$: Write lock granted\n";
    $db{$key} = $value;
    $db->sync;	# to flush
    sleep 10;

    flock(DB_FH, LOCK_UN);
    undef $db;
    untie %db;
    close(DB_FH);
    print "$$: Updated db to $key=$value\n";

Sharing databases with C applications

The vast majority of problems that are reported in this area boil down to the fact that C strings are NULL terminated, whilst Perl strings are not.

Here is a real example. Netscape 2.0 keeps a record of the locations you visit along with the time you last visited them in a DB_HASH database. This is usually stored in the file ~/.netscape/history.db. The key field in the database is the location string and the value field is the time the location was last visited stored as a 4 byte binary value.

If you haven't already guessed, the location string is stored with a terminating NULL. This means you need to be careful when accessing the database.

Here is a snippet of code that is loosely based on Tom Christiansen's ggh script (available from your nearest CPAN archive in authors/id/TOMC/scripts/nshist.gz).

    use strict ;
    use DB_File ;
    use Fcntl ;

    use vars qw( $dotdir $HISTORY %hist_db $href $binary_time $date ) ;
    $dotdir = $ENV{HOME} || $ENV{LOGNAME};

    $HISTORY = "$dotdir/.netscape/history.db";

    tie %hist_db, 'DB_File', $HISTORY
        or die "Cannot open $HISTORY: $!\n" ;;

    # Dump the complete database
    while ( ($href, $binary_time) = each %hist_db ) {

        # remove the terminating NULL
        $href =~ s/\x00$// ;

        # convert the binary time into a user friendly string
        $date = localtime unpack("V", $binary_time);
        print "$date $href\n" ;
    }

    # check for the existence of a specific key
    # remember to add the NULL
    if ( $binary_time = $hist_db{"http://mox.perl.com/\x00"} ) {
        $date = localtime unpack("V", $binary_time) ;
        print "Last visited mox.perl.com on $date\n" ;
    }
    else {
        print "Never visited mox.perl.com\n"
    }

    untie %hist_db ;

The untie gotcha

Even if you don't currently make use of the API interface, it is still worth reading it.

Here is an example which illustrates the problem from a DB_File perspective:

    use DB_File ;
    use Fcntl ;

    my %x ;
    my $X ;

    $X = tie %x, 'DB_File', 'tst.fil' , O_RDWR|O_TRUNC
        or die "Cannot tie first time: $!" ;

    $x{123} = 456 ;

    untie %x ;

    tie %x, 'DB_File', 'tst.fil' , O_RDWR|O_CREAT
        or die "Cannot tie second time: $!" ;

    untie %x ;

When run, the script will produce this error message:

    Cannot tie second time: Invalid argument at bad.file line 14.

Although the error message above refers to the second tie statement in the script, the source of the problem is really with the untie statement that precedes it.

Having read the perltie manpage you will probably have already guessed that the error is caused by the extra copy of the tied object stored in $X. If you haven't, then the problem boils down to the fact that the DB_File destructor, DESTROY, will not be called until all references to the tied object are destroyed. Both the tied variable, %x, and $X above hold a reference to the object. The call to untie will destroy the first, but $X still holds a valid reference, so the destructor will not get called and the database file tst.fil will remain open. The fact that Berkeley DB then reports the attempt to open a database that is alreday open via the catch-all ``Invalid argument'' doesn't help.

If you run the script with the -w flag the error message becomes:

    untie attempted while 1 inner references still exist at bad.file line 12.
    Cannot tie second time: Invalid argument at bad.file line 14.

which pinpoints the real problem. Finally the script can now be modified to fix the original problem by destroying the API object before the untie:

    ...
    $x{123} = 456 ;

    undef $X ;
    untie %x ;

    $X = tie %x, 'DB_File', 'tst.fil' , O_RDWR|O_CREAT
    ...

COMMON QUESTIONS

Why is there Perl source in my database?

This happens because Berkeley DB uses dynamic memory to allocate buffers which will subsequently be written to the database file. Being dynamic, the memory could have been used for anything before DB malloced it. As Berkeley DB doesn't clear the memory once it has been allocated, the unused portions will contain random junk. In the case where a Perl script gets written to the database, the random junk will correspond to an area of dynamic memory that happened to be used during the compilation of the script.

Unless you don't like the possibility of there being part of your Perl scripts embedded in a database file, this is nothing to worry about.

How do I store complex data structures with DB_File?

Check out the MLDBM module, available on CPAN in the directory modules/by-module/MLDBM.

What does "Invalid Argument" mean?

Here are a couple of possibilities:

1. Attempting to reopen a database without closing it.

2. Using the O_WRONLY flag.

What does "Bareword 'DB_File' not allowed" mean?

    use strict ;
    use DB_File ;
    use vars qw(%x) ;
    tie %x, DB_File, "filename" ;

Running it produces the error in question:

    Bareword "DB_File" not allowed while "strict subs" in use 

To get around the error, place the word DB_File in either single or double quotes, like this:

    tie %x, "DB_File", "filename" ;

Although it might seem like a real pain, it is really worth the effort of having a use strict in all your scripts.

HISTORY

1. 1 First Release.

2. 2 When DB_File is opening a database file it no longer terminates the process if dbopen returned an error. This allows file protection errors to be caught at run time. Thanks to Judith Grass <grass@cybercash.com> for spotting the bug.

3. 3 Added prototype support for multiple btree compare callbacks.

4. 0 DB_File has been in use for over a year. To reflect that, the version number has been incremented to 1.0.

   Added complete support for multiple concurrent callbacks.

   Using the push method on an empty list didn't work properly. This has been fixed.

5. 01 Fixed a core dump problem with SunOS.

   The return value from TIEHASH wasn't set to NULL when dbopen returned an error.

6. 02 Merged OS/2 specific code into DB_File.xs

   Removed some redundant code in DB_File.xs.

   Documentation update.

   Allow negative subscripts with RECNO interface.

   Changed the default flags from O_RDWR to O_CREAT|O_RDWR.

   The example code which showed how to lock a database needed a call to sync added. Without it the resultant database file was empty.

   Added get_dup method.

7. 03 Documentation update.

   DB_File now imports the constants (O_RDWR, O_CREAT etc.) from Fcntl automatically.

   The standard hash function exists is now supported.

   Modified the behavior of get_dup. When it returns an associative array, the value is the count of the number of matching BTREE values.

8. 04 Minor documentation changes.

   Fixed a bug in hash_cb. Patches supplied by Dave Hammen, <hammen@gothamcity.jsc.nasa.gov >.

   Fixed a bug with the constructors for DB_File::HASHINFO, DB_File::BTREEINFO and DB_File::RECNOINFO. Also tidied up the constructors to make them -w clean.

   Reworked part of the test harness to be more locale friendly.

9. 05 Made all scripts in the documentation strict and -w clean.

   Added logic to DB_File.xs to allow the module to be built after Perl is installed.

10. 06 Minor namespace cleanup: Localized PrintBtree.

11. 07 Fixed bug with RECNO, where bval wasn't defaulting to ``\n''.

12. 08 Documented operation of bval.

13. 09 Minor bug fix in DB_File::HASHINFO, DB_File::RECNOINFO and DB_File::BTREEINFO.

    Changed default mode to 0666.

14. 10 Fixed fd method so that it still returns -1 for in-memory files when db 1.86 is used.

15. 11 Documented the untie gotcha.

BUGS

I am sure there are bugs in the code. If you do find any, or can suggest any enhancements, I would welcome your comments.

AVAILABILITY

Berkeley DB is available at your nearest CPAN archive (see CPAN for a list) in src/misc/db.1.85.tar.gz, or via the host ftp.cs.berkeley.edu in /ucb/4bsd/db.tar.gz. Alternatively, check out the Berkeley DB home page at http://www.bostic.com/db. It is not under the GPL.

If you are running IRIX, then get Berkeley DB from http://reality.sgi.com/ariel. It has the patches necessary to compile properly on IRIX 5.3.

As of January 1997, version 1.86 of Berkeley DB is available from the Berkeley DB home page. Although this release does fix a number of bugs that were present in 1.85 you should be aware of the following information (taken from the Berkeley DB home page) before you consider using it:

    DB version 1.86 includes a new implementation of the hash access
    method that fixes a variety of hashing problems found in DB version
    1.85. We are making it available as an interim solution until DB
    2.0 is available.

    PLEASE NOTE: the underlying file format for the hash access method
    changed between version 1.85 and version 1.86, so you will have to
    dump and reload all of your databases to convert from version 1.85
    to version 1.86. If you do not absolutely require the fixes from
    version 1.86, we strongly urge you to wait until DB 2.0 is released
    before upgrading from 1.85.  

SEE ALSO

AUTHOR