NAME

Catmandu::Fix - a Catmandu class used for data transformations

SYNOPSIS

    # From the command line

    $ catmandu convert JSON --fix 'add_field(foo,bar)' < data.json
    $ catmandu convert YAML --fix 'upcase(job); remove_field(test)' < data.yml
    $ catmandu convert CSV  --fix 'sort_field(tags)' < data.csv
    $ catmandu run /tmp/myfixes.txt
    $ catmandu convert OAI --url http://biblio.ugent.be/oai --fix /tmp/myfixes.txt

    # From Perl

    use Catmandu;

    my $fixer = Catmandu->fixer('upcase(job)','remove_field(test)');
    my $fixer = Catmandu->fixer('/tmp/myfixes.txt');

    # Convert data
    my $arr      = $fixer->fix([ ... ]);
    my $hash     = $fixer->fix({ ... });
    my $importer = Catmandu->importer('YAML', file => 'data.yml');
    my $fixed_importer = $fixer->fix($importer);

    # Inline fixes
    use Catmandu::Fix::upcase       as => 'my_upcase';
    use Catmandu::Fix::remove_field as => 'my_remove';

    my $hash = { 'job' => 'librarian' , deep => { nested => '1'} };

    my_upcase($hash,'job');
    my_remove($hash,'deep.nested');

DESCRIPTION

A Catmandu::Fix is a Perl package that can transform data. These packages are used for easy data manipulation by non programmers. The main intention is to use fixes on the command line or in Fix scripts. A small DSL language is available to execute many Fix command on a stream of data.

When a fix argument is given to a Catmandu::Importer, Catmandu::Exporter or Catmandu::Store then the transformations are executed on every item in the stream.

FIX LANGUAGE

A Fix script is a collection of one or more Fix commands. The fixes are executed on every record in the dataset. If this command is executed on the command line:

    $ catmandu convert JSON --fix 'upcase(title); add_field(deep.nested.field,1)' < data.json

then all the title fields will be upcased and a new deeply nested field will be added:

    { "title":"foo" }
    { "title":"bar" }

becomes:

    { "title":"FOO" , "deep":{"nested":{"field":1}} }
    { "title":"BAR" , "deep":{"nested":{"field":1}} }

Using the command line, Fix commands need a semicolon (;) as separator. All these commands can also be written into a Fix script where semicolons are not required:

    $ catmandu convert JSON --fix script.fix < data.json

where script.fix contains:

    upcase(title)
    add_field(deep.nested.field,1)

Conditionals can be used to provide the logic when to execute fixes:

    if exists(deep.nested.field)
        add_field(nested,"ok!")
    end

    unless all_match(title,"PERL")
        add_field(is_perl,"noooo")
    end

Binds are used to manipulate the context in which Fixes are executed. E.g. execute a fix on every item in a list:

     # 'demo' is an array of hashes
     do list(path:demo)
        add_field(foo,bar)
     end

To delete records from a stream of data the reject Fix can be used:

    reject()           #  Reject all in the stream

    if exists(foo)
        reject()       # Reject records that contain a 'foo' field
    end

    reject exists(foo) # Reject records that contain a 'foo' field

The opposite of reject is select:

    select()           # Keep all records in the stream

    select exists(foo) # Keep only the records that contain a 'foo' field

Comments in Fix scripts are all lines (or parts of a line) that start with a hash (#):

    # This is ignored
    add_field(test,123)  # This is also a comment

PATHS

Most of the Fix commandsuse paths to point to values in a data record. E.g. 'foo.2.bar' is a key 'bar' which is the 3-rd value of the key 'foo'.

A special case is when you want to point to all items in an array. In this case the wildcard '*' can be used. E.g. 'foo.*' points to all the items in the 'foo' array.

For array values there are special wildcards available:

 * $append   - Add a new item at the end of an array
 * $prepend  - Add a new item at the start of an array
 * $first    - Syntactic sugar for index '0' (the head of the array)
 * $last     - Syntactic sugar for index '-1' (the tail of the array)

E.g.

 # Create { mods => { titleInfo => [ { 'title' => 'a title' }] } };
 add_field('mods.titleInfo.$append.title', 'a title');

 # Create { mods => { titleInfo => [ { 'title' => 'a title' } , { 'title' => 'another title' }] } };
 add_field('mods.titleInfo.$append.title', 'another title');

 # Create { mods => { titleInfo => [ { 'title' => 'foo' } , { 'title' => 'another title' }] } };
 add_field('mods.titleInfo.$first.title', 'foo');

 # Create { mods => { titleInfo => [ { 'title' => 'foo' } , { 'title' => 'bar' }] } };
 add_field('mods.titleInfo.$last.title', 'bar');

PERL API

The following is a list of methods available when including Catmandu::Fix as part of a Perl program.

new(fixes => [ FIX , ...])

Create a new Catmandu::Fix which will execute every FIX into a consecutive order. A FIX can be the name of a Catmandu::Fix::* routine, or the path to a plain text file containing all the fixes to be executed.

fix(HASH)

Execute all the fixes on a HASH. Returns the fixed HASH.

fix(ARRAY)

Execute all the fixes on every element in the ARRAY. Returns an ARRAY of fixes.

fix(Catmandu::Iterator)

Execute all the fixes on every item in an Catmandu::Iterator. Returns a (lazy) iterator on all the fixes.

fix(sub {})

Executes all the fixes on a generator function. Returns a new generator with fixed data.

log

Return the current logger. See Catmandu for activating the logger in your main code.

CODING

One can extend the Fix language by creating own custom-made fixes. Two methods are available to create an own Fix function:

  * Quick and easy: create a class that implements a C<fix> method.
  * Advanced: create a class that emits Perl code that will be evaled by the Fix module.

Both methods will be explained shortly.

Quick and easy

A Fix function is a Perl class in the Catmandu::Fix namespace that implements a fix method. The fix methods accepts a Perl hash as input and returns a (fixed) Perl hash as output. As an example, the code belows implements the meow Fix which inserts a 'meow' field with value 'purrrrr'.

    package Catmandu::Fix::meow;

    use Moo;

    sub fix {
        my ($self,$data) = @_;
        $data->{meow} = 'purrrrr';
        $data;
    }

    1;

Given this Perl class, the following fix statement can be used in your application:

    # Will add 'meow' = 'purrrrr' to the data
    meow()

Use the quick and easy method when your fixes are not dependent on reading or writing data from/to a JSON path. Your Perl classes need to implement their own logic to read or write data into the given Perl hash.

Fix arguments are passed as arguments to the new function of the Perl class. As in

    # In the fix file...
    meow('test123', -count => 4)

    # ...will be translated into this pseudo code
    my $fix = Catmandu::Fix::meow->new('test123', '-count', 4);

Using Moo these arguments can be catched with Catmandu::Fix::Has package:

    package Catmandu::Fix::meow;

    use Catmandu::Sane;
    use Moo;
    use Catmandu::Fix::Has;

    has msg   => (fix_arg => 1); # required parameter 1
    has count => (fix_opt => 1, default => sub { 4 }); # optional parameter 'count' with default value 4

    sub fix {
        my ($self,$data) = @_;
        $data->{meow} = $self->msg x $self->count;
        $data;
    }

    1;

Using this code the fix statement can be used like:

    # Will add 'meow' = 'purrpurrpurrpurr'
    meow('purr', -count => 4)

Advanced

The advanced method is required when one needs to read or write values from/to deeply nested JSON paths. One could parse JSON paths using the quick and easy Perl class above, but this would require a lot of inefficient for-while loops. The advanced method emits Perl code that gets compiled. This compiled code is evaled against all Perl hashes in the unput.The best way to learn this method is by inspecting some example Fix commands.

To ease the implementation of Fixed that emit Perl code some helper methods are created. Many Fix functions require a transformation of one or more values on a JSON Path. The Catmandu::Fix::SimpleGetValue provides an easy way to create such as script. In the example below we'll set the value at a JSON Path to 'purrrrr':

    package Catmandu::Fix::purrrrr;

    use Catmandu::Sane;
    use Moo;
    use Catmandu::Fix::Has;

    has path => (fix_arg => 1);

    with 'Catmandu::Fix::SimpleGetValue';

    sub emit_value {
        my ($self, $var, $fixer) = @_;
        "${var} = 'purrrrr';";
    }

    1;

Run this command as:

    # Set the value(s) of an existing path to 'purrr'
    purrrrr(my.deep.nested.path)
    purrrrr(all.my.values.*)

Notice how the emit_value of the Catmandu::Fix::purrrrr package returns Perl code and doesn't operate directy on the Perl data. The parameter $var contains only the name of a temporary variable that will hold the value of the JSON path after compiling the code into Perl.

Use Catmandu::Fix::Has to add more arguments to this fix:

    package Catmandu::Fix::purrrrr;

    use Catmandu::Sane;
    use Moo;
    use Catmandu::Fix::Has;

    has path => (fix_arg => 1);
    has msg  => (fix_opt => 1 , default => sub { 'purrrrr' });

    with 'Catmandu::Fix::SimpleGetValue';

    sub emit_value {
        my ($self, $var, $fixer) = @_;
        my $msg = $fixer->emit_string($self->msg);
        "${var} = ${msg};";
    }

    1;