NAME

Data::Stag - Structured Tags datastructures

SYNOPSIS

# PROCEDURAL USAGE
use Data::Stag qw(:all);
$doc = stag_parse($file);
@persons = stag_findnode($doc, "person");
foreach $p (@persons) {
  printf "%s, %s phone: %s\n",
    stag_sget($p, "family_name"),
    stag_sget($p, "given_name"),
    stag_sget($p, "phone_no"),
  ;
} 

# OO USAGE
use Data::Stag;
$doc = Data::Stag->new->parse($file);
@persons = $doc->findnode("person");
foreach $p (@person) {
  printf "%s, %s phone:%s\n",
    $p->sget("family_name"),
    $p->sget("given_name"),
    $p->sget("phone_no"),
  ;
}

DESCRIPTION

This module is for manipulating data as recursively nested tag/value pairs (Structured TAGs or Simple Tree AGgreggates). These datastructures can be represented as nested arrays, which have the advantage of being native to perl. A simple example is shown below:

[ person=> [  [ family_name => $family_name ],
              [ given_name  => $given_name  ],
              [ phone_no    => $phone_no    ] ] ],

Data::Stag uses a subset of XML for import and export. This means the module can also be used as a general XML parser/writer (with certain caveats).

The above set of structured tags can be represented in XML as

<person>
  <family_name>...</family_name>
  <given_name>...</given_name>
  <phone_no>...</phone_no>
</person>

Querying is performed by passing functions, for example:

# get all people in dataset with name starting 'A'
@persons = 
  $document->where('person',
                   sub {shift->sget('family_name') =~ /^A/});

One of the things that marks this module out against other XML modules is this emphasis on a functional approach as opposed to an OO approach (it may appeal to Lisp programmers).

PROCEDURAL VS OBJECT ORIENTED USAGE

Depending on your preference, this module can be used a set of procedural subroutine calls, or as method calls upon Data::Stag objects, or both.

In procedural mode, all the subroutine calls are prefixed "stag_" to avoid namespace clashes. The following two calls are equivalent:

stag_findnode($doc, "person");
$doc->findnode("person");

In object mode, you can treat any tree element as if it is an object with automatically defined methods for getting/setting the tag values.

USE OF XML

Nested arrays can be imported and exported as XML, as well as other formats. XML can be slurped into memory all at once (using less memory than an equivalent DOM tree), or a simplified SAX style event handling model can be used. Similarly, data can be exported all at once, or as a series of events.

Although this module can be used as a general XML tool, it is intended primarily as a tool for manipulating complex data using nested tag/value pairs.

By using a simpler subset of XML that can be treated as equivalent to a basic data tree structure, we can write simpler, cleaner code. This simplicity comes at a price - this module is not very suitable for XML with attributes or mixed content.

All attributes are turned into elements. This means that it will not round-trip a piece of xml with attributes in it. For some applications this is acceptable, for others it is not.

Mixed content cannot be represented in a simple tree format, so this is also expanded.

The following piece of XML

<paragraph id="1">
  example of <bold>mixed</bold>content
</paragraph>

gets parsed as if it were actually:

<paragraph>
  <paragraph-id>1</paragraph-id>
  <paragraph-text>example of</paragraph-text>
  <bold>mixed</bold>
  <paragraph-text>content</paragraph-text>
</paragraph>

This module is more suited to dealing with complex datamodels than dealing with marked up text

It can also be used as part of a SAX-style event generation / handling framework - see Data::Stag::Base

Because nested arrays are native to perl, we can specify an XML datastructure directly in perl without going through multiple object calls.

For example, instead of the lengthy

$obj->startTag("record");
$obj->startTag("field1");
$obj->characters("foo");
$obj->endTag("field1");
$obj->startTag("field2");
$obj->characters("bar");
$obj->endTag("field2");
$obj->end("record");

We can instead write

$struct = [ record => [
            [ field1 => 'foo'],
            [ field2 => 'bar']]];

If this appeals to you, then maybe this module is for you.

PARSING

parsing out subsections of a tree and changing sub-elements

use Data::Stag qw(:all);
my $tree = stag_from('xml', $xmlfile);
my ($subtree) = stag_findnode($tree, $element);
stag_set($element, $sub_element, $new_val);
print stag_xml($subtree);

OBJECT ORIENTED

the same can be done in a more OO fashion

use Data::Stag qw(:all);
my $tree = Data::Stag->from('xml', $xmlfile);
my ($subtree) = $tree->findnode($element);
$element->set($sub_element, $new_val);
print $subtree->xml;

IN A STREAM

use Data::Stag::XMLParser;
use MyTransform;      # inherits from Data::Stag::Base
my $p = Data::Stag::XMLParser->new;
my $h = MyTransform->new;   # create a handler
$p->handler($h);
$p->parse($xmlfile);

The above can be simplified like this:

use Data::Stag;
use MyTransform;      # inherits from Data::Stag::Base
my $h = MyTransform->new;
Data::Stag->new->parse(-file=>$xmlfile, -handler=>$h);

see Data::Stag::Base for writing handlers

See the Stag website at http://stag.sourceforge.net for more examples.

STRUCTURED TAGS TREE DATA STRUCTURE

A tree of structured tags is represented as a recursively nested array, the elements of the array represent nodes in the tree.

A node is a name/data pair, that can represent tags and values. A node is represented using a reference to an array, where the first element of the array is the tagname, or element, and the second element is the data

This can be visualised as a box:

+-----------+
|Name | Data|
+-----------+

In perl, we represent this pair as a reference to an array

[ Name => $Data ]

The Data can either be a list of child nodes (subtrees), or a data value.

The terminal nodes (leafs of the tree) contain data values; this is represented in perl using primitive scalars.

For example:

[ Name => 'Fred' ]

For non-terminal nodes, the Data is a reference to an array, where each element of the the array is a new node.

+-----------+
|Name | Data|
+-----------+
        |||   +-----------+
        ||+-->|Name | Data|
        ||    +-----------+
        ||    
        ||    +-----------+
        |+--->|Name | Data|
        |     +-----------+
        |     
        |     +-----------+
        +---->|Name | Data|
              +-----------+

In perl this would be:

[ Name => [
            [Name1 => $Data1],
            [Name2 => $Data2],
            [Name3 => $Data3],
          ]
];

The extra level of nesting is required to be able to store any node in the tree using a single variable. This representation has lots of advantages over others, eg hashes and mixed hash/array structures.

MANIPULATION AND QUERYING

The following example is taken from molecular biology; we have a list of species (mouse, human, fly) and a list of genes found in that species. These are cross-referenced by an identifier called tax_id. We can do a relational-style natural join on this identifier, as follows -

use Data::Stag qw(:all);
my $tree =
[ 'db' => [
  [ 'species_set' => [
    [ 'species' => [
      [ 'common_name' => 'house mouse' ],
      [ 'binomial' => 'Mus musculus' ],
      [ 'tax_id' => '10090' ]]],
    [ 'species' => [
      [ 'common_name' => 'fruit fly' ],
      [ 'binomial' => 'Drosophila melanogaster' ],
      [ 'tax_id' => '7227' ]]],
    [ 'species' => [
      [ 'common_name' => 'human' ],
      [ 'binomial' => 'Homo sapiens' ],
      [ 'tax_id' => '9606' ]]]]],
  [ 'gene_set' => [
    [ 'gene' => [
      [ 'symbol' => 'HGNC' ],
      [ 'tax_id' => '9606' ],
      [ 'phenotype' => 'Hemochromatosis' ],
      [ 'phenotype' => 'Porphyria variegata' ],
      [ 'GO_term' => 'iron homeostasis' ],
      [ 'map' => '6p21.3' ]]],
    [ 'gene' => [
      [ 'symbol' => 'Hfe' ],
      [ 'synonym' => 'MR2' ],
      [ 'tax_id' => '10090' ],
      [ 'GO_term' => 'integral membrane protein' ],
      [ 'map' => '13 A2-A4' ]]]]]]];

# natural join of species and gene parts of tree,
# based on 'tax_id' element
my ($gene_set) = $tree->findnode("gene_set");
my ($species_set) = $tree->findnode("species_set");
$gene_set->njoin("gene", "tax_id", $species_set);
print $gene_set->xml;

# find all genes starting with H in human
my @genes =
  $gene_set->where('gene',
                   sub { my $g = shift;
                         $g->get_symbol =~ /^H/ &&
                         $g->findval("common_name") eq ('human')});

S-Expression (Lisp) representation

The data represented using this module can be represented as Lisp-style S-Expressions.

See Data::Stag::SxprParser and Data::Stag::SxprWriter

If we execute this line

print $tree->sxpr;

The following S-Expression will be printed:

'(db
  (species_set
    (species
      (common_name "house mouse")
      (binomial "Mus musculus")
      (tax_id "10090"))
    (species
      (common_name "fruit fly")
      (binomial "Drosophila melanogaster")
      (tax_id "7227"))
    (species
      (common_name "human")
      (binomial "Homo sapiens")
      (tax_id "9606")))
  (gene_set
    (gene
      (symbol "HGNC")
      (tax_id "9606")
      (phenotype "Hemochromatosis")
      (phenotype "Porphyria variegata")
      (GO_term "iron homeostasis")
      (map
        (cytological
          (chromosome "6")
          (band "p21.3"))))
    (gene
      (symbol "Hfe")
      (synonym "MR2")
      (tax_id "10090")
      (GO_term "integral membrane protein")))
  (similarity_set
    (pair
      (symbol "HGNC")
      (symbol "Hfe"))
    (pair
      (symbol "WNT3A")
      (symbol "Wnt3a"))))

TIPS FOR EMACS USERS AND LISP PROGRAMMERS

If you use emacs, you can save this as a file with the ".el" suffix and get syntax highlighting for editing this file. Quotes around the terminal node data items are optional.

If you know emacs lisp or any other lisp, this also turns out to be a very nice language for manipulating these datastructures. Try copying and pasting the above s-expression to the emacs scratch buffer and playing with it!

I think this module turns out to be a very nice way using my two favourite lnaguages, lisp and perl together.

INDENTED TEXT REPRESENTATION

Data::Stag has its own text format for writing data trees. Again, this is only possible because we are working with a subset of XML (no attributes, no mixed elements). The data structure above can be written as follows -

db:
  species_set:
    species:
      common_name: house mouse
      binomial: Mus musculus
      tax_id: 10090
    species:
      common_name: fruit fly
      binomial: Drosophila melanogaster
      tax_id: 7227
    species:
      common_name: human
      binomial: Homo sapiens
      tax_id: 9606
  gene_set:
    gene:
      symbol: HGNC
      tax_id: 9606
      phenotype: Hemochromatosis
      phenotype: Porphyria variegata
      GO_term: iron homeostasis
      map: 6p21.3
    gene:
      symbol: Hfe
      synonym: MR2
      tax_id: 10090
      GO_term: integral membrane protein
      map: 13 A2-A4
  similarity_set:
    pair:
      symbol: HGNC
      symbol: Hfe
    pair:
      symbol: WNT3A
      symbol: Wnt3a

See Data::Stag::ITextParser and Data::Stag::ITextWriter

NESTED ARRAY SPECIFICATION II

To avoid excessive square bracket usage, you can specify a structure like this:

use Data::Stag qw(:all);

*N = \&stag_new;
my $tree =
  N(top=>[
          N('personset'=>[
                          N('person'=>[
                                       N('name'=>'davey'),
                                       N('address'=>'here'),
                                       N('description'=>[
                                                         N('hair'=>'green'),
                                                         N('eyes'=>'two'),
                                                         N('teeth'=>5),
                                                        ]
                                        ),
                                       N('pets'=>[
                                                  N('petname'=>'igor'),
                                                  N('petname'=>'ginger'),
                                                 ]
                                        ),
                                                                        
                                      ],
                           ),
                          N('person'=>[
                                       N('name'=>'shuggy'),
                                       N('address'=>'there'),
                                       N('description'=>[
                                                         N('hair'=>'red'),
                                                         N('eyes'=>'three'),
                                                         N('teeth'=>1),
                                                        ]
                                        ),
                                       N('pets'=>[
                                                  N('petname'=>'thud'),
                                                  N('petname'=>'spud'),
                                                 ]
                                        ),
                                      ]
                           ),
                         ]
           ),
          N('animalset'=>[
                          N('animal'=>[
                                       N('name'=>'igor'),
                                       N('class'=>'rat'),
                                       N('description'=>[
                                                         N('fur'=>'white'),
                                                         N('eyes'=>'red'),
                                                         N('teeth'=>50),
                                                        ],
                                        ),
                                      ],
                           ),
                         ]
           ),

         ]
   );

# find all people
my @persons = stag_findnode($tree, 'person');

# write xml for all red haired people
foreach my $p (@persons) {
  print stag_xml($p)
    if stag_tmatch("hair", "red");
} ;

# find all people called shuggy
my @p =
  stag_qmatch($tree, 
              "person",
              "name",
              "shuggy");

NODES AS DATA OBJECTS

As well as the methods listed below, a node can be treated as if it is a data object of a class determined by the element.

For example, the following are equivalent.

$node->get_name;
$node->get('name');

$node->set_name('fred');
$node->set('name', 'fred');

This is really just syntactic sugar. The autoloaded methods are not checked against any schema, although this may be added in future.

One addition slated for a future release is the ability to give particular elements certain behaviour, and allow inheritance and all that kind of thing.

fullname: $obj->given_name . ' ' . $obj->family_name;

Although it is the module authors preference to avoid this kind of OO paradigm, and instead enforce a cleaner seperation of code from data, utilising a more functional style of programming.

METHODS

All method calls are also available as procedural subroutine calls; unless otherwise noted, the subroutine call is the same as the method call, but with the string stag_ prefixed to the method name. The first argument should be a Data::Stag datastructure.

To import all subroutines into the current namespace, use this idiom:

use Data::Stag qw(:all);

If you wish to use this module procedurally, and you are too lazy to prefix all calls with stag_, use this idiom:

use Data::Stag qw(:lazy);

MNEMONICS

Most method calls also have a handy short mnemonic. Use of these is optional. Software engineering types prefer longer names, in the belief that this leads to clearer code. Hacker types prefer shorter names, as this requires less keystrokes, and leads to a more compact representation of the code. It is expected that if you do use this module, then its usage will be fairly ubiquitous within your code, and the mnemonics will become familiar, much like the qw and s/ operators in perl. As always with perl, the decision is yours.

INITIALIZATION METHODS

new

  Title: new

   Args: element str, data ANY
Returns: Data::Stag node
Example: $node = stag_new();
Example: $node = Data::Stag->new;
Example: $node = Data::Stag->new(person => [[name=>$n], [phone=>$p]]);

creates a new instance of a Data::Stag node

nodify

  Title: nodify

   Args: data array-reference
Returns: Data::Stag node
Example: $node = stag_nodify([person => [[name=>$n], [phone=>$p]]]);

turns a perl array reference into a Data::Stag node.

similar to new

parse

  Title: parse

   Args: file str, [format str], [handler obj]
Returns: Data::Stag node
Example: $node = stag_parse($fn);
Example: $node = Data::Stag->parse(-file=>$fn, -handler=>$myhandler);

slurps a file or string into a Data::Stag node structure. Will guess the format from the suffix if it is not given.

The format can also be the name of a parsing module, or an actual parser object

from

  Title: from

   Args: format str, source str
Returns: Data::Stag node
Example: $node = stag_from('xml', $fn);
Example: $node = stag_from('xmlstr', q[<top><x>1</x></top>]);
Example: $node = Data::Stag->from($parser, $fn);

Similar to parse

slurps a file or string into a Data::Stag node structure.

The format can also be the name of a parsing module, or an actual parser object

unflatten

  Title: unflatten

   Args: data array
Returns: Data::Stag node
Example: $node = stag_unflatten(person=>[name=>$n, phone=>$p, address=>[street=>$s, city=>$c]]);

Creates a node structure from a semi-flattened representation, in which children of a node are represented as a flat list of data rather than a list of array references.

This means a structure can be specified as:

person=>[name=>$n,
         phone=>$p, 
         address=>[street=>$s, 
                   city=>$c]]

Instead of:

[person=>[ [name=>$n],
           [phone=>$p], 
           [address=>[ [street=>$s], 
                       [city=>$c] ] ]
         ]
]

The former gets converted into the latter for the internal representation

RECURSIVE SEARCHING

findnode (fn)

  Title: findnode
Synonym: fn

   Args: element str
Returns: node[]
Example: @persons = stag_findnode($struct, 'person');
Example: @persons = $struct->findnode('person');

recursively searches tree for all elements of the given type, and returns all nodes found.

findval (fv)

  Title: findval
Synonym: fv

   Args: element str
Returns: ANY[] or ANY
Example: @names = stag_findval($struct, 'name');
Example: @names = $struct->findval('name');
Example: $firstname = $struct->findval('name');

recursively searches tree for all elements of the given type, and returns all data values found. the data values could be primitive scalars or nodes.

sfindval (sfv)

  Title: sfindval
Synonym: sfv

   Args: element str
Returns: ANY
Example: $name = stag_sfindval($struct, 'name');
Example: $name = $struct->sfindval('name');

as findval, but returns the first value found

findvallist (fvl)

  Title: findvallist
Synonym: fvl

   Args: element str[]
Returns: ANY[]
Example: ($name, $phone) = stag_findvallist($personstruct, 'name', 'phone');
Example: ($name, $phone) = $personstruct->findvallist('name', 'phone');

recursively searches tree for all elements in the list

DEPRECATED?

DATA ACCESSOR METHODS

these allow getting and setting of elements directly underneath the current one

get (g)

  Title: get
Synonym: g

   Args: element str
 Return: ANY
Example: $name = $person->get('name');
Example: @phone_nos = $person->get('phone_no');

gets the data value of an element for any node

the examples above would work on a data structure like this:

[person => [ [name => 'fred'],
             [phone_no => '1-800-111-2222'],
             [phone_no => '1-415-555-5555']]]

will return an array or single value depending on the context

[equivalent to findval(), except that only direct children (as opposed to all descendents) are checked]

sget (sg)

  Title: sget
Synonym: sg

   Args: element str
 Return: ANY
Example: $name = $person->get('name');
Example: $phone = $person->get('phone_no');

as get but always returns a single value

[equivalent to sfindval(), except that only direct children (as opposed to all descendents) are checked]

getl (gl getlist)

  Title: gl
Synonym: getl
Synonym: getlist

   Args: element str[]
 Return: ANY[]
Example: ($name, @phone) = $person->get('name', 'phone_no');

returns the data values for a list of sub-elements of a node

[equivalent to findvallist(), except that only direct children (as opposed to all descendents) are checked]

getn (gn getnode)

  Title: getn
Synonym: gn
Synonym: getnode

   Args: element str
 Return: node[]
Example: $namestruct = $person->getn('name');
Example: @pstructs = $person->getn('phone_no');

as get but returns the whole node rather than just the data valie

[equivalent to findnode(), except that only direct children (as opposed to all descendents) are checked]

set (s)

  Title: set
Synonym: s

   Args: element str, datavalue ANY
 Return: ANY
Example: $person->set('name', 'fred');
Example: $person->set('phone_no', $cellphone, $homephone);

sets the data value of an element for any node. if the element is multivalued, all the old values will be replaced with the new ones specified.

ordering will be preserved, unless the element specified does not exist, in which case, the new tag/value pair will be placed at the end.

note that if the datavalue is a non-terminal node as opposed to a primitive value, then you have to do it like this:

$address = Data::Stag->new(address=>[
                                     [address_line=>"221B Baker Street"],
                                     [city=>"London"],
                                     [country=>"Great Britain"]]);
($person) = $data->qmatch("name", "Sherlock Holmes");
$person->set("address", $address->data);

unset (u)

  Title: unset
Synonym: u

   Args: element str, datavalue ANY
 Return: ANY
Example: $person->unset('name');
Example: $person->unset('phone_no');

prunes all nodes of the specified element from the current node

add (a)

  Title: add
Synonym: a

   Args: element str, datavalue ANY[]
 Return: ANY
Example: $person->add('phone_no', $cellphone, $homephone);

adds a datavalue or list of datavalues. appends if already existing, creates new element value pairs if not already existing.

element (e name)

  Title: element
Synonym: e
Synonym: name

   Args:
 Return: element str
Example: $element = $struct->element

returns the element name of the current node

kids (k children)

  Title: kids
Synonym: k
Synonym: children

   Args:
 Return: ANY or ANY[]
Example: @nodes = $person->kids
Example: $name = $namestruct->kids

returns the data value(s) of the current node; if it is a terminal node, returns a single value which is the data. if it is non-terminal, returns an array of nodes

addkid (ak addchild)

  Title: addkid
Synonym: ak
Synonym: addchild

   Args: kid node
 Return: ANY
Example: $person->addkid('job', $job);

adds a new child node to a non-terminal node, after all the existing child nodes

subnodes

  Title: subnodes

   Args: 
 Return: ANY[]
Example: @nodes = $person->subnodes

returns the non-terminal data value(s) of the current node;

QUERYING AND ADVANCED DATA MANIPULATION

njoin (j)

  Title: njoin
Synonym: j
Synonym: nj

   Args: element str
 Return: undef

does a relational style natural join - see previous example in this doc

qmatch (qm)

  Title: qmatch
Synonym: qm

   Args: return-element str, match-element str, match-value str
 Return: node[]
Example: @persons = $s->qmatch('name', 'fred');

queries the node tree for all elements that satisfy the specified key=val match

tmatch (tm)

  Title: tmatch
Synonym: tm

   Args: element str, value str
 Return: bool
Example: @persons = grep {$_->tmatch('name', 'fred')} @persons

returns true if the the value of the specified element matches

tmatchhash (tmh)

  Title: tmatchhash
Synonym: tmh

   Args: match hashref
 Return: bool
Example: @persons = grep {$_->tmatchhash({name=>'fred', hair_colour=>'green'})} @persons

returns true if the node matches a set of constraints, specified as hash

tmatchnode (tmn)

  Title: tmatchnode
Synonym: tmn

   Args: match node
 Return: bool
Example: @persons = grep {$_->tmatchnode([person=>[[name=>'fred'], [hair_colour=>'green']]])} @persons

returns true if the node matches a set of constraints, specified as node

cmatch (cm)

  Title: cmatch
Synonym: cm

   Args: element str, value str
 Return: bool
Example: $n_freds = $personset->cmatch('name', 'fred');

counts the number of matches

where (w)

  Title: where
Synonym: w

   Args: element str, test CODE
 Return: Node[]
Example: @rich_persons = $data->where('person', sub {shift->get_salary > 100000});

the tree is queried for all elements of the specified type that satisfy the coderef (must return a boolean)

my @rich_dog_or_cat_owners =
  $data->where('person',
               sub {my $p = shift;
                    $p->get_salary > 100000 &&
                    $p->where('pet',
                              sub {shift->get_type =~ /(dog|cat)/})});

MISCELLANEOUS METHODS

duplicate (d)

  Title: duplicate
Synonym: d

   Args:
 Return: Node
Example: $node2 = $node->duplicate;

isanode

  Title: isanode

   Args:
 Return: bool
Example: if (stag_isanode($node)) { ... }

really only useful in non OO mode...

hash

  Title: hash

   Args:
 Return: hash
Example: $h = $node->hash;

turns a tree into a hash. all data values will be arrayrefs

pairs

Title: pairs

turns a tree into a hash. all data values will be scalar (IMPORTANT: this means duplicate values will be lost)

write

  Title: write

   Args: filename str, format str[optional]
 Return:
Example: $node->write("myfile.xml");
Example: $node->write("myfile", "itext");

will try and guess the format from the extension if not specified

xml

  Title: xml

   Args: filename str, format str[optional]
 Return:
Example: $node->write("myfile.xml");
Example: $node->write("myfile", "itext");


   Args:
 Return: xml str
Example: print $node->xml;

XML METHODS

sax

  Title: sax

   Args: saxhandler SAX-CLASS
 Return:
Example: $node->sax($mysaxhandler);

turns a tree into a series of SAX events

xpath (xp tree2xpath)

  Title: xpath
Synonym: xp
Synonym: tree2xpath

   Args:
 Return: xpath object
Example: $xp = $node->xpath; $q = $xp->find($xpathquerystr);

xpquery (xpq xpathquery)

  Title: xpquery
Synonym: xpq
Synonym: xpathquery

   Args: xpathquery str
 Return: Node[]
Example: @nodes = $node->xqp($xpathquerystr);

BUGS

none known so far, possibly quite a few undocumented features!

Not a bug, but the underlying default datastructure of nested arrays is more heavyweight than it needs to be. More lightweight implementations are possible. Some time I will write a C implementation.

WEBSITE

http://stag.sourceforge.net

WEBSITE

http://stag.sourceforge.net

AUTHOR

Chris Mungall <cjm@fruitfly.org>

COPYRIGHT

Copyright (c) 2002 Chris Mungall

This module is free software. You may distribute this module under the same terms as perl itself