stdlib_hashmap_wrappers and stdlib_hashmaps:local:
:depth: 2
A hash map (hash table) is a data structure that maps keys to values. It uses a hash function to compute a hash code from the key that serves as an index into a linear array of slots (buckets) from which the desired value can be extracted. Each key ideally maps to a unique slot, but most hash functions are imperfect and can map multiple keys to the same slot resulting in collisions. Hash maps differ in how they deal with such collisions. This document discusses the hash maps in the Fortran Standard Library.
The Fortran Standard Library is distributed under the MIT License. However components of the library should be evaluated as to whether they are compatible with the MIT License. The current hash maps were inspired by an implementation of David Chase. While the code has been greatly modified from his implementation, he has give permission for the unrestricted use of his code.
The Fortran Standard Library provides two modules for the
implementation of simple hash maps. These maps only accept hash
functions with a single argument, the key, and yield a 32 bit
hash code. The modules will need to be modified if it is desired to
use hash functions with a different API. The two modules are:
stdlib_hashmap_wrappers, and stdlib_hashmaps corresponding to the
files: stdlib_hashmap_wrappers.f90, and stdlib_hashmaps.f90
The module stdlib_hashmap_wrappers provides types and procedures for
use by stdlib_hashmaps. It provides an
interface to the 32 bit hash functions of the Standard Library module,
stdlib_hash_32bit, and provides wrappers to some of the
hash functions so that they no longer need to be supplied seeds. It
also defines the key_type derived type. The key_type is used to
define keys that, in turn, are used to identify the data entered into
a hash map.
The module stdlib_hashmaps defines the API for a parent datatype,
hashmap_type and two extensions of that hash map type:
chaining_hashmap_type and open_hashmap_type.
The hashmap_type defines the Application Programmers
Interface (API) for the procedures used by its two extensions. It
explicitly defines five non-overridable procedures. It also defines
the interfaces for eleven deferred procedures. It does not define the
finalization routines for the two extension types, or one routine
provided by the open_hashmap_type.
The chaining_hashmap_type uses separate chaining with linked
lists to deal with hash index collisions. In separate chaining the
colliding indices are handled by using linked lists with their roots
at the hash index. The chaining_hashmap_type procedures are
implemented in the module stdlib_hashmap_chaining corresponding
to the file, stdlib_hashmap_chaining.f90.
The open_hashmap_type
uses linear open addressing to deal with hash index collisions. In
linear open addressing the colliding indices are
handled by searching from the initial hash index in increasing
steps of one (modulo the hash map size) for an open map slot.
The open_hashmap_type procedures are implemented in the submodule
stdlib_hashmap_open corresponding to the file
stdlib_hashmap_open.f90.
The maps use powers of two for their slot sizes, so that the function,
fibonacci_hash, can
be used to map the hash codes to indices in the map. This is
expected to be more efficient than prime number mapping using a
modulo operation, and reduces the requirement that the hash
function need to do a good job randomizing its lower order bits.
They do require a good randomizing hash method for good performance.
Both adjust the map size to reduce collisions, based on
the ratio of the number of hash map probes to the number of subroutine
calls.
Wile the maps make extensive use of pointers internally, a private
finalization subroutine avoids memory leaks.
The maps can take entry keys of type key_type, and other data (also
commonly known as values, as in key value pairs) in any scalar type.
The maps allow the addition, removal, and lookup of entries, and the
inclusion of data in addition to the entry key.
stdlib_hashmap_wrappers moduleThe stdlib_hashmap_wrappers module provides data types to
represent keys and associated data stored in a module, but is also, a
wrapper for the stdlib_hash_32bit module. It allows
direct access to the stdlib_hash_32bit procedures:
fibonacci_hash, fnv_1_hasher, fnv_1a_hasher; and provides
wrapper functions, seeded_nmhash32_hasher,
seeded_nmhash32x_hasher, and seeded_water_hasher to the hash
functions: nmhash32, nmhash32x, and water_hash, respectively. It
defines an interface, hasher_fun, compatible with the hash functions
that take a non-scalar key. It defines one integer constant used
as a kind value,int_hash. It also defines two types, key_type and
other_type, and associated procedures, for storing and manipulating
keys and their associated data.
stdlib_hashmap_wrappers’s constant, int_hashThe constant int_hash is used to define the integer kind value for
the returned hash codes and variables used to access them. It
currently is imported from stdlib_hash_32bit where it has the
value, int32.
stdlib_hashmap_wrappers’ module’s derived typesThe stdlib_hashmap_wrappers defines key_type which is intended to
be used for the search keys of hash tables. The tye is opaque.
The current representation is as follows
type :: key_type
private
integer(int8), allocatable :: value(:)
end type key_type
The module also defines six procedures for those types: copy_key,
equal_keys, free_key, get, set, and one operator, ==,
for use by the hash maps to manipulate or inquire of components of
those types.
stdlib_hashmap_wrappers proceduresThe stdlib_hashmap_wrappers module provides procedures in
several categories: procedures to manipulate data of the key_type;
and 32 bit hash functions for keys. The procedures in each category
are listed below. It also provides an operator to compare two key
type values for equality.
Procedures to manipulate key_type data:
copy_key( key_in, key_out ) - Copies the contents of the key,
key_in, to contents of the key, key_out.
get( key, value ) - extracts the contents of key into value,
an int8 array, int32 array, or character string.
free_key( key ) - frees the memory in key.
set( key, value ) - sets the content of key to value.
Supported key types are int8 array, int32 array, and character
string.
Procedures to hash keys to 32 bit integers:
fnv_1_hasher( key ) - hashes a key using the FNV-1 algorithm.
fnv_1a_hasher( key ) - hashes a key using the FNV-1a algorithm.
seeded_nmhash32_hasher( key ) - hashes a key using the nmhash32
algorithm.
seeded_nmhash32x_hasher( key ) - hashes a key using the nmhash32x
algorithm.
seeded_water_hasher( key ) - hashes a key using the waterhash
algorithm.
Operator to compare two key_type values for equality
key1 == key2 - compares key1 with key2 for equalitystdlib_hashmap_wrappers procedurescopy_key - Returns a copy of the keyExperimental
Returns a copy of an input of type key_type.
call {f:subr}copy_key ( old_key, new_key )
Subroutine.
old_key: shall be a scalar expression of type key_type. It
is an intent(in) argument.
new_key: shall be a scalar variable of type key_type. It
is an intent(out) argument.
```{literalinclude} ../../example/hashmaps/example_hashmaps_copy_key.f90 :language: fortran
#### `fibonacci_hash` - maps an integer to a smaller number of bits
##### Status
Experimental
##### Description
`fibonacci_hash` is just a re-export of the function of the same name
implemented in
[`stdlib_hash_32bit`](/stdlib/doc/specs/stdlib_hash_procedures.html).
It reduces the value of a 32 bit integer to a smaller number of bits.
#### `fnv_1_hasher`- calculates a hash code from a key
##### Status
Experimental
##### Description
Calculates a 32 bit hash code from an input of type `key_type`.
##### Syntax
`code = ` {f:func}`fnv_1_hasher` `( key )`
##### Class
Pure function
##### Argument
`key`: Shall be a scalar expression of type `key_type`.
It is an `intent(in)` argument.
##### Result character
The result is a scalar integer of kind `int32`.
##### Result value
The result is a hash code created using the FNV-1 algorithm.
##### Note
`fnv_1_hasher` is an implementation of the original FNV-1 hash code of
Glenn Fowler, Landon Curt Noll, and Phong Vo.
This code is relatively fast on short keys, and is small enough that
it will often be retained in the instruction cache if hashing is
intermittent.
As a result it should give good performance for typical hash map
applications.
This code does not pass any of the SMHasher tests, but the resulting
degradation in performance due to its larger number of collisions is
expected to be minor compared to its faster hashing rate.
##### Example
```{literalinclude} ../../example/hashmaps/example_hashmaps_fnv_1_hasher.f90
:language: fortran
fnv_1a_hasher- calculates a hash code from a keyExperimental
Calculates a 32 bit hash code from an input of type key_type.
code = fnv_1a_hasher ( key )
Pure function
key: Shall be a scalar expression of type key_type.
It is an intent(in) argument.
The result is a scalar integer of kind int32.
The result is a hash code created using the FNV-1a algorithm.
fnv_1a_hasher is an implementation of the original FNV-1A hash code
of Glenn Fowler, Landon Curt Noll, and Phong Vo.
This code is relatively fast on short keys, and is small enough that
it will often be retained in the instruction cache if hashing is
intermittent.
As a result it should give good performance for typical hash map
applications.
This code does not pass any of the SMHasher tests, but the resulting
degradation in performance due to its larger number of collisions is
expected to be minor compared to its faster hashing rate.
```{literalinclude} ../../example/hashmaps/example_hashmaps_fnv_1a_hasher.f90 :language: fortran
#### `free_key` - frees the memory associated with a key
##### Status
Experimental
##### Description
Deallocates the memory associated with a variable of type
`key_type`.
##### Syntax
`call ` {f:subr}`free_key` `( key )`
##### Class
Subroutine.
##### Argument
`key`: shall be a scalar variable of type `key_type`. It
is an `intent(out)` argument.
##### Example
```{literalinclude} ../../example/hashmaps/example_hashmaps_free_key.f90
:language: fortran
get - extracts the data from a derived typeExperimental
Extracts the data from a key_type and stores it in the
variable value.
call {f:subr}get ( key, value )
Subroutine.
key: shall be a scalar expression of type key_type. It
is an intent(in) argument.
value: shall be an allocatable default character string variable,
or an allocatable vector variable of type integer and kind int8 or
int32.
```{literalinclude} ../../example/hashmaps/example_hashmaps_get.f90 :language: fortran
#### `hasher_fun`- serves as a function prototype.
##### Status
Experimental
##### Description
Serves as a prototype for hashing functions with a single, `key`,
argument of type `key_type` returning an `int32` hash value.
##### Syntax
`type(` `hasher_fun` `), pointer :: fun_pointer`
##### Class
Pure function prototype
##### Argument
`key`: Shall be a rank one array expression of type `integer(int8)`.
It is an `intent(in)` argument.
##### Result character
The result is a scalar integer of kind `int32`.
##### Result value
The result is a hash code.
##### Note
`hasher_fun` is a prototype for defining dummy arguments and function
pointers intended for use as a hash function for the hash maps.
##### Example
```{literalinclude} ../../example/hashmaps/example_hashmaps_hasher_fun.f90
:language: fortran
operator(==) - Compares two keys for equalityExperimental
Returns .true. if two keys are equal, and .false. otherwise.
test = key1 == key2
Pure operator.
key1: shall be a scalar expression of type key_type. It
is an intent(in) argument.
key2: shall be a scalar expression of type key_type. It
is an intent(in) argument.
The result is a value of type default logical.
The result is .true. if the keys are equal, otherwise .falss..
```{literalinclude} ../../example/hashmaps/example_hashmaps_equal_keys.f90
:language: fortran
#### `seeded_nmhash32_hasher`- calculates a hash code from a key
##### Status
Experimental
##### Description
Calculates a 32 bit hash code from an input of type `key_type`.
##### Syntax
`code = ` `seeded_nmhash32_hasher` `( key )`
##### Class
Pure function
##### Argument
`key`: Shall be a scalar expression of type `key_type`.
It is an `intent(in)` argument.
##### Result character
The result is a scalar integer of kind `int32`.
##### Result value
The result is a hash code created using the `nmhash32` algorithm.
##### Note
`seeded_nmhash32_hasher` is a wrapper to the `NMHASH32_HASH` of the
module `stdlib_hash_32bit`, which supplies a fixed seed
to the wrapped function. `NMHASH32` is an implementation of the
`nmhash32` hash code of James Z. M. Gao.
This code has good, but not great, performance on long keys, poorer
performance on short keys.
As a result it should give fair performance for typical hash map
applications.
This code passes the SMHasher tests.
##### Example
```{literalinclude} ../../example/hashmaps/example_hashmaps_seeded_nmhash32_hasher.f90
:language: fortran
seeded_nmhash32x_hasher- calculates a hash code from a keyExperimental
Calculates a 32 bit hash code from an input of type key_type.
code = seeded_nmhash32x_hasher ( key )
Pure function
key: Shall be a scalar expression of type key_type.
It is an intent(in) argument.
The result is a scalar integer of kind int32.
The result is a hash code created using the nmhash32x algorithm.
seeded_nmhash32x_hasher is a wrapper to the nmhash32x_hash of the
module stdlib_hash_32bit, which supplies a fixed seed
to the wrapped function. nmhash32x is an implementation of the
nmhash32x hash code of James Z. M. Gao.
This code has good, but not great, performance on long keys, poorer
performance on short keys.
As a result it should give fair performance for typical hash map
applications.
This code passes the SMHasher tests.
```{literalinclude} ../../example/hashmaps/example_hashmaps_seeded_nmhash32x_hasher.f90 :language: fortran
#### `seeded_water_hasher`- calculates a hash code from a key
##### Status
Experimental
##### Description
Calculates a 32 bit hash code from an input of type `key_type`.
##### Syntax
`code = ` `seeded_water_hasher` `( key )`
##### Class
Pure function
##### Argument
`key`: Shall be a scalar expression of type `key_type`.
It is an `intent(in)` argument.
##### Result character
The result is a scalar integer of kind `int32`.
##### Result value
The result is a hash code created using the `waterhash` algorithm.
##### Note
`seeded_water_hasher` is a wrapper to the `water_hash` of the
module `stdlib_hash_32bit`, which supplies a fixed seed
to the wrapped function. `water_hash` is an implementation of the
`waterhash` hash code of Tommy Ettinger.
This code has excellent performance on long keys, and good performance
on short keys.
As a result it should give reasonable performance for typical hash
table applications.
This code passes the SMHasher tests.
##### Example
```{literalinclude} ../../example/hashmaps/example_hashmaps_seeded_water_hasher.f90
:language: fortran
set - places the data in a derived typeExperimental
Places the data from value in a key_type.
call {f:subr}set ( key, value )
Subroutine.
key: shall be a scalar variable of type key_type. It
is an intent(out) argument.
value: shall be a default character string scalar expression,
or a vector expression of type integerand kind int8 or int32.
It is an intent(in) argument.
Values of types other than a scalar default character or and
int8 or int32 vector can be used as the basis of a key by transferring the
value to an int8 vector.
```{literalinclude} ../../example/hashmaps/example_hashmaps_set.f90 :language: fortran
## The `stdlib_hashmaps` module
The `stdlib_hashmaps` module defines three public data types,
associated procedures and constants that implement two simple hash map
types using separate chaining hashing and open addressing hashing. The
derived type `hashmap_type` is the parent type to its two
extensions: `chaining_hashmap_type` and `open_hashmap_type`.
The extension types provide
procedures to manipulate the structure of a hash map object:
`init`, `map_entry`, `rehash`, `remove`, and
`set_other_data`. They also provide procedures to inquire about
entries in the hash map: `get_other_data`, and
`key_test`. Finally they provide procedures to inquire about the
overall structure and performance of the hash map object:`calls`,
`entries`, `get_other_data`, `loading`, `slots`, and
`total_depth`. The module also defines a number of public constants:
`probe_factor`, `load_factor`, `map_probe_factor`, `default_bits`,
`max_bits`, `int_calls`, `int_depth`, `int_index`,
`int_probes`, `success`, `alloc_fault`, and `array_size_error`.
Generic key interfaces for `key_test`, `map_entry`, `get_other_data`,
`remove`, and `set_other_data` are povided so that the supported types
of `int8` arrays, `int32` arrays and `character` scalars can be used in the
key field as well as the base `key` type. So for `key_test`,
`key_key_test` specifies key type for the key field, `int8_key_test` is `int8`
for the key field and so on. Procedures other than `key_key_test` will call
the `set` function to generate a key type and pass to `key_key_test`.
### The `stdlib_hashmaps` module's public constants
The module defines several categories of public constants. Some are
used to parameterize the empirical slot expansion code. Others
parameterize the slots table size. Some are used to define
integer kind values for different applications. Finally, some are used
to report errors or success.
The constants `probe_factor`, and `map_probe_factor` are used to
parameterize the slot expansion code used to determine when in a
in a procedure call the number
of slots need to be increased to decrease the search path for an
entry. The constant `probe_factor` is used to determine when
the ratio of the number of map probes to map calls is too large and
the slots need expansion. The constant `map_probe_factor` is used to
determine when inserting a new entry the ratio of the number of map
probes to map calls is too large and the slots need expansion.
The constants `default_bits`, and
`max_bits` are used to parameterize the table's slots size. The
`default_bits` constant defines the default initial number of slots
with a current value of 6 resulting in an initial `2**6 == 64`
slots. This may optionally be overridden on hash map creation. The
`max_bits` parameter sets the maximum table size as `2**max_bits` with
a default value for `max_bits` of 30. The table will not work for a
slots size greater than `2**30`.
The constants `int_calls`, `int_depth`, `int_index`, and `int_probes`
are used to define integer kind values for various contexts. The
number of calls are reported and stored in entities of kind
`int_calls`. Currently `int_calls` has the value of `int64`. The
total depth, the number of inquiries needed to access all elements
of the table, is reported and stored in entities of kind
`int_depth`. Currently `int_depth` has the value of `int64`. The
number of entries in the table, is reported and stored in entities of
kind `int_index`. Currently `int_index` has the value of `int32`.
The number of probes, hash map enquiries, are reported and stored in
entities of kind `int_probes`. Currently `int_probes` has the value of
`int64`.
The constant `load_factor` is only used by the `open_hashmap_type`. It
specifies the maximum fraction of the available slots that may be
filled before expansion occurs. The current `load_factor = 0.5625` so
the current implementation of `open_hashmap_type` can only hold a
little more than `2**29` entries.
Finally the error codes `success`, `alloc_fault`, and
`array_size_error` are used to report the error status of certain
procedure calls. The `succes` code indicates that no problems were
found. The `alloc_fault` code indicates that a memory allocation
failed. Finally the `array_size_error` indicates that on table
creation `slots_bits` is less than `default_bits` or
greater than `max_bits`.
### The `stdlib_hashmaps` module's derived types
The `stdlib_hashmaps` module defines three public derived types and
seven private types used in the implementation of the public
types. The public types are the abstract `hashmap_type` and its
extensions: `chaining_hashmap_type` and `open_hashmap_type`. The three
private derived types, `chaining_map_entry_type`,
`chaining_map_entry_ptr`, and `chaining_map_entry_pool` are used in
the implementation of the `chaining_hashmap_type` public type. The
four private derived types, `open_map_entry_type`,
`open_map_entry_list`, `open_map_entry_ptr`, and `open_map_entry_pool`
are used in the implementation of the `open_hashmap_type` public
type. Each of these types are described below.
#### The `hashmap_type` abstract type
The `hashmap_type` abstract type serves as the parent type for the two
types `chaining_hashmap_type` and `open_hashmap_type`. It defines
eight private components:
* `call_count` - the number of procedure calls on the map;
* `nbits` - the number of bits used to address the slots;
* `num_entries` - the number of entries in the map;
* `num_free` - the number of entries in the free list of removed
entries;
* `probe_count` - the number of map probes since the last resizing or
initialization;
* `total_probes` - the number of probes of the map up to the last
resizing or initialization; and
* `hasher` - a pointer to the hash function used by the map.
* `initialized` - track if map has been initialized
It also defines five non-overridable procedures:
* `calls` - returns the number of procedure calls on the map;
* `entries` - returns the number of entries in the map;
* `map_probes` - returns the number of map probes since
initialization;
* `num_slots` - returns the number of slots in the map; and
* `slots_bits` - returns the number of bits used to address the slots;
and ten deferred procedures:
* `get_all_keys` - gets all the keys contained in a map;
* `get_other_data` - gets the value associated with a key;
* `init` - initializes the hash map;
* `key_test` - returns a logical flag indicating whether the key is
defined in the map.
* `loading` - returns the ratio of the number of entries to the number
of slots;
* `map_entry` - inserts a key and optionally a corresponding value into
the map;
* `rehash` - rehashes the map with the provided hash function;
* `remove` - removes the entry associated wit the key;
* `set_other_data` - replaces the value associated with a key;
* `total_depth` - returns the number of probes needed to address all
the entries in the map;
The type's definition is below:
```fortran
type, abstract :: hashmap_type
private
integer(int_calls) :: call_count = 0
integer(int_calls) :: probe_count = 0
integer(int_calls) :: total_probes = 0
integer(int_index) :: num_entries = 0
integer(int_index) :: num_free = 0
integer(int32) :: nbits = default_bits
procedure(hasher_fun), pointer, nopass :: hasher => fnv_1_hasher
contains
procedure, non_overridable, pass(map) :: calls
procedure, non_overridable, pass(map) :: entries
procedure, non_overridable, pass(map) :: map_probes
procedure, non_overridable, pass(map) :: num_slots
procedure, non_overridable, pass(map) :: slots_bits
procedure(get_all_keys), deferred, pass(map) :: get_all_keys
procedure(init_map), deferred, pass(map) :: init
procedure(loading), deferred, pass(map) :: loading
procedure(rehash_map), deferred, pass(map) :: rehash
procedure(total_depth), deferred, pass(map) :: total_depth
!! Generic interfaces for key types.
procedure(key_key_test), deferred, pass(map) :: key_key_test
procedure, non_overridable, pass(map) :: int8_key_test
procedure, non_overridable, pass(map) :: int32_key_test
procedure, non_overridable, pass(map) :: char_key_test
procedure(key_map_entry), deferred, pass(map) :: key_map_entry
procedure, non_overridable, pass(map) :: int8_map_entry
procedure, non_overridable, pass(map) :: int32_map_entry
procedure, non_overridable, pass(map) :: char_map_entry
procedure(key_get_other_data), deferred, pass(map) :: key_get_other_data
procedure, non_overridable, pass(map) :: int8_get_other_data
procedure, non_overridable, pass(map) :: int32_get_other_data
procedure, non_overridable, pass(map) :: char_get_other_data
procedure(key_remove_entry), deferred, pass(map) :: key_remove_entry
procedure, non_overridable, pass(map) :: int8_remove_entry
procedure, non_overridable, pass(map) :: int32_remove_entry
procedure, non_overridable, pass(map) :: char_remove_entry
procedure(key_set_other_data), deferred, pass(map) :: key_set_other_data
procedure, non_overridable, pass(map) :: int8_set_other_data
procedure, non_overridable, pass(map) :: int32_set_other_data
procedure, non_overridable, pass(map) :: char_set_other_data
generic, public :: key_test => key_key_test, int8_key_test, int32_key_test, char_key_test
generic, public :: map_entry => key_map_entry, int8_map_entry, int32_map_entry, char_map_entry
generic, public :: get_other_data => key_get_other_data, int8_get_other_data, int32_get_other_data, char_get_other_data
generic, public :: remove => key_remove_entry, int8_remove_entry, int32_remove_entry, char_remove_entry
generic, public :: set_other_data => key_set_other_data, int8_set_other_data, int32_set_other_data, char_set_other_data
end type hashmap_type
chaining_map_entry_type derived typeEntities of the type chaining_map_entry_type are used to define
a linked list structure that stores the
key, its other data, the hash of the key, and the resulting index into
the inverse table. The type’s definition is below:
type :: chaining_map_entry_type ! Chaining hash map entry type
private
integer(int_hash) :: hash_val ! Full hash value
type(key_type) :: key ! The entry's key
class(*), allocatable :: other ! Other entry data
integer(int_index) :: index ! Index into inverse table
type(chaining_map_entry_type), pointer :: &
next => null() ! Next bucket
end type chaining_map_entry_type
Currently the int_hash and int_index have the value of int32.
chaining_map_entry_ptr derived typeThe type chaining_map_entry_ptr is used to define the elements of
the hash map that are either empty or link to the linked lists
containing the elements of the table. The type’s definition is below:
type chaining_map_entry_ptr ! Wrapper for a pointer to a chaining
! map entry type object
type(chaining_map_entry_type), pointer :: target => null()
end type chaining_map_entry_ptr
chaining_map_entry_pool derived typeThe type chaining_map_entry_pool is used to implement a pool of
allocated chaining_map_entry_type elements to save on allocation
costs. The type’s definition is below:
type :: chaining_map_entry_pool
! Type implementing a pool of allocated
! `chaining_map_entry_type` objects
private
! Index of next bucket
integer(int_index) :: next = 0
type(chaining_map_entry_type), allocatable :: more_map_entries(:)
type(chaining_map_entry_pool), pointer :: lastpool => null()
end type chaining_map_entry_pool
chaining_hashmap_type derived typeThe chaining_hashmap_type derived type extends the hashmap_type to
implements a separate chaining hash map. In addition to the components
of the hashmap_type it provides the four components:
cache - a pool of chaining_map_entry_pool objects used to reduce
allocation costs;
free_list - a free list of map entries;
inverse - an array of chaining_map_entry_ptr bucket lists
(inverses) storing entries at fixed locations once
entered; and
slots - an array of bucket lists serving as the hash map.
It also implements all of the deferred procedures of the
hashmap_type and a finalizer for its maps. The type’s definition is
as follows:
type, extends(hashmap_type) :: chaining_hashmap_type
private
type(chaining_map_entry_pool), pointer :: cache => null()
type(chaining_map_entry_type), pointer :: free_list => null()
type(chaining_map_entry_ptr), allocatable :: inverse(:)
type(chaining_map_entry_ptr), allocatable :: slots(:)
contains
procedure :: get_all_keys => get_all_chaining_keys
procedure :: key_get_other_data => get_other_chaining_data
procedure :: init => init_chaining_map
procedure :: loading => chaining_loading
procedure :: key_map_entry => map_chain_entry
procedure :: rehash => rehash_chaining_map
procedure :: key_remove_entry => remove_chaining_entry
procedure :: key_set_other_data => set_other_chaining_data
procedure :: total_depth => total_chaining_depth
procedure :: key_key_test => chaining_key_test
final :: free_chaining_map
end type chaining_hashmap_type
open_map_entry_type derived typeEntities of the type open_map_entry_type are used to define
a linked list structure that stores the
key, its other data, the hash of the key, and the resulting index into
the inverse table. The type’s definition is below:
type :: open_map_entry_type ! Open hash map entry type
private
integer(int_hash) :: hash_val ! Full hash value
type(key_type) :: key ! The entry's key
class(*), allocatable :: other ! Other entry data
integer(int_index) :: index ! Index into inverse table
end type open_map_entry_type
Currently int_hash and int_index have the value of int32.
open_map_entry_ptr derived typeThe type open_map_entry_ptr is used to define the elements of
the hash map that are either empty or link to the linked lists
containing the elements of the table. The type’s definition is below:
type open_map_entry_ptr ! Wrapper for a pointer to a open
! map entry type object
type(open_map_entry_type), pointer :: target => null()
end type open_map_entry_ptr
open_hashmap_type derived typeThe open_hashmap_type derived type extends the hashmap_type to
implement an open addressing hash map. In addition to the components
of the hashmap_type it provides the four components:
cache - a pool of open_map_entry_pool objects used to reduce
allocation costs;
free_list - a free list of map entries;
index_mask - an and mask used in linear addressing;
inverse - an array of open_map_entry_ptr bucket lists
(inverses) storing entries at fixed locations once
entered; and
slots - an array of bucket lists serving as the hash map.
It also implements all of the deferred procedures of the
hashmap_type and a finalizer for its maps. The type’s definition is
as follows:
type, extends(hashmap_type) :: open_hashmap_type
private
integer(int_index) :: index_mask = 2_int_index**default_bits-1
type(open_map_entry_pool), pointer :: cache => null()
type(open_map_entry_list), pointer :: free_list => null()
type(open_map_entry_ptr), allocatable :: inverse(:)
integer(int_index), allocatable :: slots(:)
contains
procedure :: get_all_keys => get_all_open_keys
procedure :: key_get_other_data => get_other_open_data
procedure :: init => init_open_map
procedure :: loading => open_loading
procedure :: key_map_entry => map_open_entry
procedure :: rehash => rehash_open_map
procedure :: key_remove_entry => remove_open_entry
procedure :: key_set_other_data => set_other_open_data
procedure :: total_depth => total_open_depth
procedure :: key_key_test => open_key_test
final :: free_open_map
end type open_hashmap_type
stdlib_hashmap proceduresThe stdlib_hashmap module provides procedures in
several categories: a procedure to initialize the map; a procedure to
modify the structure of a map; procedures to modify the content of a
map; procedures to report on the content of a map; and procedures
to report on the structure of the map. The procedures in each category
are listed below.
Procedure to initialize a chaining hash map:
map % init( [hasher, slots_bits, status] ) - Routine
to initialize a chaining hash map.Procedure to modify the structure of a map:
map % rehash( hasher ) - Routine to change the hash function
for a map.Procedures to modify the content of a map:
map % map_entry( key[, other, conflict] ) - Inserts an entry into the
hash map.
map % remove( key[, existed] ) - Remove the entry, if any,
associated with the key.
map % set_other_data( key, other[, exists] ) - Change the value
associated with the key.
Procedures to report the content of a map:
map % get_all_keys( all_keys ) - Returns all the keys
contained in the map;
map % get_other_data( key, other[, exists] ) - Returns the value
associated with the key;
map % key_test( key, present) - Returns a flag indicating whether
the key is present in the map.
Procedures to report on the structure of the map:
map % calls() - the number of subroutine calls on the hash map.
map % entries()- the number of entries in a hash map.
map % loading() - the number of entries relative to the number of
slots in a hash map.
map % map_probes() - the total number of table probes on a hash
map.
map % slots() - Returns the number of allocated slots in a hash
map.
map % total_depth() - Returns the total number of one’s based
offsets of slot entries from their slot index
stdlib_hashmaps procedurescalls - Returns the number of calls on the hash mapExperimental
Returns the number of procedure calls on a hash map.
value = map % calls(bound) ()
Pure function
map (pass) - shall be an expression of class hashmap_type.
It is an intent(in) argument.
The result will be an integer of kind int_calls.
The result will be the number of procedure calls on the hash map.
```{literalinclude} ../../example/hashmaps/example_hashmaps_calls.f90 :language: fortran
#### `entries` - Returns the number of entries in the hash map
##### Status
Experimental
##### Description
Returns the number of entries in a hash map.
##### Syntax
`value = map % ` `entries(bound)` `()`
##### Class
Pure function
##### Argument
`map` (pass) - shall be an expression of class `hashmap_type`.
It is an `intent(in)` argument.
##### Result character
The result will be an integer of kind `int_index`.
##### Result value
The result will be the number of entries in the hash map.
##### Example
```{literalinclude} ../../example/hashmaps/example_hashmaps_entries.f90
:language: fortran
get_all_keys - Returns all the keys contained in a mapExperimental
Returns all the keys contained in a map.
call map % get_all_keys(bound) ( all_keys )
Subroutine
map (pass): shall be a scalar variable of class
chaining_hashmap_type or open_hashmap_type. It is an
intent(in) argument. It will be
the hash map used to store and access the other data.
all_keys: shall be a rank-1 allocatable array of type key_type.
It is an intent(out) argument.
```{literalinclude} ../../example/hashmaps/example_hashmaps_get_all_keys.f90 :language: fortran
#### `get_other_data` - Returns other data associated with the `key`
##### Status
Experimental
##### Description
Returns the value associated with the `key`,
##### Syntax
`value = map % ` `get_other_data(bound)` `( key, other [, exists] )`
##### Class
Subroutine
##### Arguments
`map` (pass): shall be a scalar variable of class
`chaining_hashmap_type` or `open_hashmap_type`. It is an
`intent(inout)` argument. It will be
the hash map used to store and access the other data.
`key`: shall be a of type `key_type` scalar, `character` scalar, `int8` array
or `int32` array. It is an `intent(in)` argument.
`other`: shall be a allocatable unlimited polymorphic scalar.
(class(*), allocatable) It is an `intent(out)` argument.
It is the value associated with the `key`.
`exists` (optional): shall be a variable of type logical. It is an
`intent(out)` argument. If `.true.` an entry with the given `key`
exists in the map and `other` is defined. If `.false.` `other` is
undefined.
##### Example
The following is an example of the retrieval of other data
associated with a `key`:
```{literalinclude} ../../example/hashmaps/example_hashmaps_get_other_data.f90
:language: fortran
init - initializes a hash mapExperimental
Initializes a hashmap_type object.
call map % init(bound) ( [hasher, slots_bits, status ] )
Subroutine
map (pass): shall be a scalar variable of class
chaining_hashmap_type or open_hashmap_type. It is an
intent(out) argument. It will
be a hash map used to store and access the entries.
hasher: (optional): shall be a procedure with interface hash_fun.
It is an intent(in) argument. It is the procedure to be used to
generate the hashes for the table from the keys of the entries.
Defaults to fnv_1_hasher if not provided.
slots_bits (optional): shall be a scalar default integer
expression. It is an intent(in) argument. The initial number of
slots in the table will be 2**slots_bits.
slots_bits shall be a positive default integer less than
max_bits, otherwise processing stops with an informative
error code.
If slots_bits is absent then the effective value for slots_bits
is default_bits.
status (optional): shall be a scalar integer variable of kind
int32. It is an intent(out) argument. On return if present it
shall have an error code value.
If map was successfully initialized then status has the value
success.
If allocation of memory for the map arrays fails then status
has the value alloc_fault.
If slot_bits < 6 or slots_bits > max_bits then status
has the value of array_size_error.
If status is absent, but status would have a value other than
success, then processing stops with an informative stop code.
```{literalinclude} ../../example/hashmaps/example_hashmaps_init.f90 :language: fortran
#### `key_test` - indicates whether `key` is present
##### Status
Experimental
##### Description
Returns a logical flag indicating whether `key` is present for an
entry in the map.
##### Syntax
`call map % ` `key_test(bound)` `( key, present )`
##### Class
Subroutine.
##### Arguments
`map` (pass): shall be a scalar variable of class
`chaining_hashmap_type` or `open_hashmap_type`.
It is an `intent(inout)` argument. It is the hash map whose entries
are examined.
`key`: shall be a of type `key_type` scalar, `character` scalar, `int8` array
or `int32` array. It is an `intent(in)` argument. It is a `key` whose
presence in the `map` is being examined.
`present`: shall be a scalar variable of type `logical`.
It is an `intent(out)` argument. It is a logical flag where
`.true.` indicates that an entry with that `key` is present in the
`map` and `.false.` indicates that no such entry is present.
##### Example
```{literalinclude} ../../example/hashmaps/example_hashmaps_key_test.f90
:language: fortran
loading - Returns the ratio of entries to slotsExperimental
Returns the ratio of the number of entries relative to the number of slots in the hash map.
value = map % loading(bound) ( )
Pure function
map (pass) - shall be an expression of class chaining_hashmap_type
or open_hashmap_type. It is an intent(in) argument.
The result will be a default real.
The result will be the ratio of the number of entries relative to the number of slots in the hash map.
```{literalinclude} ../../example/hashmaps/example_hashmaps_loading.f90 :language: fortran
#### `map_entry` - inserts an entry into the hash map
##### Status
Experimental
##### Description
Inserts an entry into the hash map if it is not already present.
##### Syntax
`call map % ` `map_entry(bound)` `( key[, other, conflict ] )`
##### Class
Subroutine
##### Arguments
`map` (pass): shall be a scalar variable of class
`chaining_hashmap_type` or `open_hashmap_type`. It
is an `intent(inout)` argument. It is the hash map to receive the
entry.
`key`: shall be a of type `key_type` scalar, `character` scalar, `int8` array
or `int32` array. It is an `intent(in)` argument. It is the key for the entry
to be placed in the table.
`other` (optional): shall be a scalar of any type, including derived types.
It is an `intent(in)` argument. If present it is the value to be
associated with the `key`.
`conflict` (optional): shall be a scalar variable of type
`logical`. It is an `intent(out)` argument. If present, a `.true.`
value indicates that an entry with the value of `key` already exists
and the entry was not entered into the map, a `.false.` value indicates
that `key` was not present in the map and the entry was added to the
map.
* If `key` is already present in `map` and the `conflict` argument has been
provided then the presence of `other` is ignored. If `conflict` has not
been provided then it routine will error stop.
##### Example
```{literalinclude} ../../example/hashmaps/example_hashmaps_map_entry.f90
:language: fortran
map_probes - returns the number of hash map probesExperimental
Returns the total number of table probes on the hash map.
result = map % map_probes(bound) ( )
Pure function
map (pass): shall be a scalar expression of class
hashmap_type. It is an intent(in)
argument. It is the hash map of interest.
The result is a scalar integer of kind int_probes.
The result is the number of probes of map since initialization or
rehashing.
```{literalinclude} ../../example/hashmaps/example_hashmaps_probes.f90 :language: fortran
#### `num_slots` - returns the number of hash map slots.
##### Status
Experimental
##### Description
Returns the total number of slots on a hash map
##### Syntax
`result = map % ` `num_slots(bound)` `( )`
##### Class
Pure function
##### Argument
`map`: shall be a scalar expression of class
`hashmap_type`. It is an `intent(in)` argument. It is the
hash map of interest.
##### Result character
The result is a scalar integer of kind `int_index`.
##### Result value
The result is the number of slots in `map`.
##### Example
```{literalinclude} ../../example/hashmaps/example_hashmaps_num_slots.f90
:language: fortran
rehash - changes the hashing functionExperimental
Changes the hashing function for the map entries to that of hasher.
call map % rehash(bound) ( hasher )
Subroutine
map (pass): shall be a scalar variable of class
chaining_hashmap_type or open_hashmap_type.
It is an intent(inout) argument. It is the hash map whose hashing
method is to be changed.
hasher: shall be a function of interface hasher_fun.
It is the hash method to be used by map.
```{literalinclude} ../../example/hashmaps/example_hashmaps_rehash.f90 :language: fortran
#### `remove` - removes an entry from the hash map
##### Status
Experimental
##### Description
Removes an entry from the hash map, `map`.
##### Syntax
`call map % ` `remove(bound)` `( key[, existed ])`
##### Class
Subroutine
##### Arguments
`map` (pass): shall be a scalar variable of class
`chaining_hashmap_type` or `open_hashmap_type`.
It is an `intent(inout)` argument. It is the hash map with the element
to be removed.
`key`: shall be a of type `key_type` scalar, `character` scalar, `int8` array
or `int32` array. It is an `intent(in)` argument. It is the `key` identifying
the entry to be removed.
`existed` (optional): shall be a scalar variable of type default
logical. It is an `intent(out)` argument. If present with the value
`.true.` the entry existed in the map before removal, if `.false.` the
entry was not present to be removed and the map is unchanged. If
absent, the procedure returns with no entry with the given key.
##### Example
```{literalinclude} ../../example/hashmaps/example_hashmaps_remove.f90
:language: fortran
set_other_data - replaces the other data for an entryExperimental
Replaces the other data in the map for the entry with the key value,
key.
call map % set_other_data(bound) ( key, other[, exists] )
Subroutine
map (pass): shall be a scalar variable of class
chaining_hashmap_type or open_hashmap_type. It
is an intent(inout) argument. It will be a hash map used to store
and access the entry’s data.
key: shall be a of type key_type scalar, character scalar, int8 array
or int32 array. It is an intent(in) argument. It is the key to the
entry whose other data is to be replaced.
other (optional): shall be a scalar of any type, including derived types.
It is an intent(in) argument. If present it is the value to be
associated with the key.
exists (optional): shall be a scalar variable of type logical.
It is an intent(out) argument. If present with the value
.true. an entry with that key existed in the map and its other
data was replaced. If exists is .false. the key did
not exist and nothing was done.
key is not already present in map and exists has not
been provided then the routine will error stop.```{literalinclude} ../../example/hashmaps/example_hashmaps_set_other_data.f90 :language: fortran
#### `slots_bits` - returns the number of bits used to address the hash map slots
##### Status
Experimental
##### Description
Returns the total number of bits used to address the hash map slots.
##### Syntax
`result = map % ` `slots_bits(bound)` `( )`
##### Class
Pure function
##### Argument
`map` (pass): shall be a scalar expression of class
`hashmap_type`. It is an `intent(in)` argument. It is the
hash map of interest.
##### Result character
The result is a scalar integer of kind `int_index`.
##### Result value
The result is the number of bits used in addressing the slots in `map`.
##### Example
```{literalinclude} ../../example/hashmaps/example_hashmaps_slots_bits.f90
:language: fortran
total_depth - returns the total depth of the hash map entriesExperimental
Returns the total number of one’s based offsets of slot entries from their slot index for a hash map
result = map % total_depth ( )
Pure function
map (pass): shall be a scalar expression of class
hashmap_type. It is an intent(in) argument. It is the
hash map of interest.
The result is a scalar integer of kind int_depth.
The result is the total number of one’s based offsets of slot entries from their slot index the map.
{literalinclude} ../../example/hashmaps/example_hashmaps_total_depth.f90
:language: fortran