NAME Hash::Ordered - A fast, pure-Perl ordered hash class VERSION version 0.010 SYNOPSIS use Hash::Ordered; my $oh = Hash::Ordered->new( a => 1 ); $oh->get( 'a' ); $oh->set( 'a' => 2 ); $oh->exists( 'a' ); $val = $oh->delete( 'a' ); @keys = $oh->keys; @vals = $oh->values; @pairs = $oh->as_list $oh->push( c => 3, d => 4 ); $oh->unshift( e => 5, f => 6 ); ( $k, $v ) = $oh->pop; ( $k, $v ) = $oh->shift; $iter = $oh->iterator; while( ( $k, $v ) = $iter->() ) { ... } $copy = $oh->clone; $subset = $oh->clone( qw/c d/ ); $reversed = $oh->clone( reverse $oh->keys ); @value_slice = $oh->values( qw/c f/ ); # qw/3 6/ @pairs_slice = $oh->as_list( qw/f e/ ); # qw/f 6 e 5/ $oh->postinc( 'a' ); # like $oh{a}++ $oh->add( 'a', 5 ); # like $oh{a} += 5 $oh->concat( 'a', 'hello' ); # like $oh{a} .= 'hello' $oh->or_equals( 'g', '23' ); # like $oh{g} ||= 23 $oh->dor_equals( 'g', '23' ); # like $oh{g} //= 23 DESCRIPTION This module implements an ordered hash, meaning that it associates keys with values like a Perl hash, but keeps the keys in a consistent order. Because it is implemented as an object and manipulated with method calls, it is much slower than a Perl hash. This is the cost of keeping order. However, compared to other ordered hash implementations, Hash::Ordered is optimized for getting and setting individual elements and is generally faster at most other tasks as well. For specific details, see Hash::Ordered::Benchmarks. METHODS new $oh = Hash::Ordered->new; $oh = Hash::Ordered->new( @pairs ); Constructs an object, with an optional list of key-value pairs. The position of a key corresponds to the first occurrence in the list, but the value will be updated if the key is seen more than once. Current API available since 0.009. clone $oh2 = $oh->clone; $oh2 = $oh->clone( @keys ); Creates a shallow copy of an ordered hash object. If no arguments are given, it produces an exact copy. If a list of keys is given, the new object includes only those keys in the given order. Keys that aren't in the original will have the value "undef". keys @keys = $oh->keys; $size = $oh->keys; In list context, returns the ordered list of keys. In scalar context, returns the number of elements. Current API available since 0.005. values @values = $oh->values; @values = $oh->values( @keys ); Returns an ordered list of values. If no arguments are given, returns the ordered values of the entire hash. If a list of keys is given, returns values in order corresponding to those keys. If a key does not exist, "undef" will be returned for that value. In scalar context, returns the number of elements. Current API available since 0.006. get $value = $oh->get("some key"); Returns the value associated with the key, or "undef" if it does not exist in the hash. set $oh->set("some key" => "some value"); Associates a value with a key and returns the value. If the key does not already exist in the hash, it will be added at the end. exists if ( $oh->exists("some key") ) { ... } Test if some key exists in the hash (without creating it). delete $value = $oh->delete("some key"); Removes a key-value pair from the hash and returns the value. clear $oh->clear; Removes all key-value pairs from the hash. Returns undef in scalar context or an empty list in list context. Current API available since 0.003. push $oh->push( one => 1, two => 2); Add a list of key-value pairs to the end of the ordered hash. If a key already exists in the hash, it will be deleted and re-inserted at the end with the new value. Returns the number of keys after the push is complete. pop ($key, $value) = $oh->pop; $value = $oh->pop; Removes and returns the last key-value pair in the ordered hash. In scalar context, only the value is returned. If the hash is empty, the returned key and value will be "undef". unshift $oh->unshift( one => 1, two => 2 ); Adds a list of key-value pairs to the beginning of the ordered hash. If a key already exists, it will be deleted and re-inserted at the beginning with the new value. Returns the number of keys after the unshift is complete. shift ($key, $value) = $oh->shift; $value = $oh->shift; Removes and returns the first key-value pair in the ordered hash. In scalar context, only the value is returned. If the hash is empty, the returned key and value will be "undef". merge $oh->merge( one => 1, two => 2 ); Merges a list of key-value pairs into the ordered hash. If a key already exists, its value is replaced. Otherwise, the key-value pair is added at the end of the hash. as_list @pairs = $oh->as_list; @pairs = $oh->as_list( @keys ); Returns an ordered list of key-value pairs. If no arguments are given, all pairs in the hash are returned. If a list of keys is given, the returned list includes only those key-value pairs in the given order. Keys that aren't in the original will have the value "undef". iterator $iter = $oh->iterator; $iter = $oh->iterator( reverse $oh->keys ); # reverse while ( my ($key,$value) = $iter->() ) { ... } Returns a code reference that returns a single key-value pair (in order) on each invocation, or the empty list if all keys are visited. If no arguments are given, the iterator walks the entire hash in order. If a list of keys is provided, the iterator walks the hash in that order. Unknown keys will return "undef". The list of keys to return is set when the iterator is generator. Keys added later will not be returned. Subsequently deleted keys will return "undef" for the value. preinc $oh->preinc($key); # like ++$hash{$key} This method is sugar for incrementing a key without having to call "set" and "get" explicitly. It returns the new value. Current API available since 0.005. postinc $oh->postinc($key); # like $hash{$key}++ This method is sugar for incrementing a key without having to call "set" and "get" explicitly. It returns the old value. Current API available since 0.005. predec $oh->predec($key); # like --$hash{$key} This method is sugar for decrementing a key without having to call "set" and "get" explicitly. It returns the new value. Current API available since 0.005. postdec $oh->postdec($key); # like $hash{$key}-- This method is sugar for decrementing a key without having to call "set" and "get" explicitly. It returns the old value. Current API available since 0.005. add $oh->add($key, $n); # like $hash{$key} += $n This method is sugar for adding a value to a key without having to call "set" and "get" explicitly. With no value to add, it is treated as "0". It returns the new value. Current API available since 0.005. subtract $oh->subtract($key, $n); # like $hash{$key} -= $n This method is sugar for subtracting a value from a key without having to call "set" and "get" explicitly. With no value to subtract, it is treated as "0". It returns the new value. Current API available since 0.005. concat $oh->concat($key, $str); # like $hash{$key} .= $str This method is sugar for concatenating a string onto the value of a key without having to call "set" and "get" explicitly. It returns the new value. If the value to append is not defined, no concatenation is done and no warning is given. Current API available since 0.005. or_equals $oh->or_equals($key, $str); # like $hash{$key} ||= $str This method is sugar for assigning to a key if the existing value is false without having to call "set" and "get" explicitly. It returns the new value. Current API available since 0.005. dor_equals $oh->dor_equals($key, $str); # like $hash{$key} //= $str This method is sugar for assigning to a key if the existing value is not defined without having to call "set" and "get" explicitly. It returns the new value. Current API available since 0.005. OVERLOADING Boolean if ( $oh ) { ... } When used in boolean context, a Hash::Ordered object is true if it has any entries and false otherwise. String say "$oh"; When used in string context, a Hash::Ordered object stringifies like typical Perl objects. E.g. "Hash::Ordered=ARRAY(0x7f815302cac0)" Current API available since 0.005. Numeric $count = 0 + $oh; When used in numeric context, a Hash::Ordered object numifies as the decimal representation of its memory address, just like typical Perl objects. E.g. 140268162536552 For the number of keys, call the "keys" method in scalar context. Current API available since 0.005. Fallback Other overload methods are derived from these three, if possible. TIED INTERFACE Using "tie" is slower than using method calls directly. But for compatibility with libraries that can only take hashes, it's available if you really need it: tie my %hash, "Hash::Ordered", @pairs; If you want to access the underlying object for method calls, use "tied": tied( %hash )->unshift( @data ); Tied hash API available since 0.005. CAVEATS Deletion and order modification with push, pop, etc. This can be expensive, as the ordered list of keys has to be updated. For small hashes with no more than 25 keys, keys are found and spliced out with linear search. As an optimization for larger hashes, the first change to the ordered list of keys will construct an index to the list of keys. Thereafter, removed keys will be marked with a "tombstone" record. Tombstones will be garbage collected whenever the number of tombstones exceeds the number of valid keys. These internal implementation details largely shouldn't concern you. The important things to note are: * The costs of efficient deletion are deferred until you need it * Deleting lots of keys will temporarily appear to leak memory until garbage collection occurs MOTIVATION For a long time, I used Tie::IxHash for ordered hashes, but I grew frustrated with things it lacked, like a cheap way to copy an IxHash object or a convenient iterator when not using the tied interface. As I looked at its implementation, it seemed more complex than I though it needed, with an extra level of indirection that slows data access. Given that frustration, I started experimenting with the simplest thing I thought could work for an ordered hash: a hash of key-value pairs and an array with key order. As I worked on this, I also started searching for other modules doing similar things. What I found fell broadly into two camps: modules based on tie (even if they offered an OO interface), and pure OO modules. They all either lacked features I deemed necessary or else seemed overly-complex in either implementation or API. Hash::Ordered attempts to find the sweet spot with simple implementation, reasonably good efficiency for most common operations, and a rich, intuitive API. After discussions with Mario Roy about the potential use of Hash::Ordered with MCE, I optimized deletion of larger hashes and provided a tied interface for compatibility. Mario's suggestions and feedback about optimization were quite valuable. Thank you, Mario! SEE ALSO This section describes other ordered-hash modules I found on CPAN. For benchmarking results, see Hash::Ordered::Benchmarks. Tie modules The following modules offer some sort of tie interface. I don't like ties, in general, because of the extra indirection involved over a direct method call. Still, you can make any tied interface into a faster OO one with "tied": tied( %tied_hash )->FETCH($key); Tie::Hash::Indexed is implemented in XS and thus seems promising if pure-Perl isn't a criterion; it generally fails tests on Perl 5.18 and above due to the hash randomization change. Despite being XS, it is slower than Hash::Ordered at everything exception creation and deletion. Tie::IxHash is probably the most well known and includes an OO API. Given the performance problems it has, "well known" is the only real reason to use it. These other modules below have very specific designs/limitations and I didn't find any of them suitable for general purpose use: * Tie::Array::AsHash — array elements split with separator; tie API only * Tie::Hash::Array — ordered alphabetically; tie API only * Tie::InsertOrderHash — ordered by insertion; tie API only * Tie::LLHash — linked-list implementation; quite slow * Tie::StoredOrderHash — ordered by last update; tie API only Other ordered hash modules Other modules stick with an object-oriented API, with a wide variety of implementation approaches. Array::AsHash is essentially an inverse implementation from Hash::Ordered. It keeps pairs in an array and uses a hash to index into the array. This indirection would already make hash-like operations slower, but the specific implementation makes it even worse, with abstractions and function calls that make getting or setting individual items up to 10x slower than Hash::Ordered. However, "Array::AsHash" takes an arrayref to initialize, which is very fast and can return the list of pairs faster, too. If you mostly create and list out very large ordered hashes and very rarely touch individual entries, I think this could be something to very cautiously consider. These other modules below have restrictions or particularly complicated implementations (often relying on "tie") and thus I didn't think any of them really suitable for use: * Array::Assign — arrays with named access; restricted keys * Array::OrdHash — overloads array/hash deref and uses internal tied data * Data::Pairs — array of key-value hashrefs; allows duplicate keys * Data::OMap — array of key-value hashrefs; no duplicate keys * Data::XHash — blessed, tied hashref with doubly-linked-list SUPPORT Bugs / Feature Requests Please report any bugs or feature requests through the issue tracker at . You will be notified automatically of any progress on your issue. Source Code This is open source software. The code repository is available for public review and contribution under the terms of the license. git clone https://github.com/dagolden/Hash-Ordered.git AUTHOR David Golden CONTRIBUTORS * Benct Philip Jonsson * David Golden * Mario Roy COPYRIGHT AND LICENSE This software is Copyright (c) 2014 by David Golden. This is free software, licensed under: The Apache License, Version 2.0, January 2004