Title

Basic hash tables

Author

Panu Kalliokoski

Status

This SRFI is currently in ``draft'' status. To see an explanation of each status that a SRFI can hold, see here. It will remain in draft status until 2005/09/09, or as amended. To provide input on this SRFI, please mailto:srfi-69@srfi.schemers.org. See instructions here to subscribe to the list. You can access previous messages via the archive of the mailing list.

Abstract

This SRFI specifies an API for basic hash tables. Hash tables are data structures that provide a mapping from some set of keys to some set of values associated to those keys. Hash tables have no intrinsic order for these associations, and allow key lookup and destructive updating in amortised constant time (when a good hash function is used).

Issues

There is no single best way to make hash tables. The tables presented in this SRFI aim at being both conceptually simple and usable for a wide variety of applications. Even though a portable implementation is provided, Scheme implementations can speed things up considerably by e.g. providing an internal hash function for symbols. Moreover, almost every Scheme implementation already has some kind of low-level hash table functionality, because that's the natural way to implement the global environment, and specifically, to provide support for string->symbol. There might be some benefit in integration between implementation-specific environment data types and the hash table API presented here; however, these issues are left open.

This SRFI does not conform to the interface of maps presented in SRFI 44. An implementation supporting both SRFI 44 and this SRFI is encouraged to provide a SRFI 44 interface to hash tables in addition to the one presented here.

Rationale

Hash tables are widely recognised as a fundamental data structure for many kinds of computational tasks. Almost every non-minimal Scheme implementation provides some kind of hash table functionality.

Alas, although somewhat similar, these hash table APIs have many differences: some trivial, like the naming of certain functions; some complex, like revealing different aspects of the internal implementation to the user; some coarse, like requiring keys to be of some specific type(s); some subtle, like requiring the user to guess the size of the hash table in advance to get optimal performance. As a result, the easiest way to write portable Scheme programs that use hash tables is to implement some kind of hash tables in the program itself, based on vectors.

The primary aim of this SRFI is to provide a simple and generic hash table API that will answer most of users' needs for basic usage of hash tables; the reference implementation just shows one way of meeting the requirements of the API.

Specification

Names defined in this SRFI:

Type constructors and predicate
make-hash-table, make-symbol-hash-table, make-string-hash-table, make-string-ci-hash-table, make-integer-hash-table, hash-table?, alist->hash-table
Reflective queries
hash-table-equivalence-function, hash-table-hash-function
Dealing with single elements
hash-table-ref, hash-table-ref/default, hash-table-set!, hash-table-delete!, hash-table-exists?, hash-table-update!, hash-table-update!/default
Dealing with the whole contents
hash-table-size, hash-table-keys, hash-table-values, hash-table-walk, hash-table-fold, hash-table->alist, hash-table-copy, hash-table-merge!
Hashing
hash, symbol-hash, string-hash, string-ci-hash, hash-by-identity

Type constructors and predicate

Procedure: make-hash-table [ equal? [ hash [ sizehint ]]] → hash-table

Create a new hash table with no associations. equal? is a predicate that should accept two keys and return a boolean telling whether they denote the same key value; it defaults to equal?.

hash is a hash function, and defaults to an appropriate hash function for the given equal? predicate (see section Hashing). However, an acceptable default is not guaranteed to be given for any equivalence predicate coarser than equal?, except for string-ci=?. A equivalence predicate c1 is coarser than a equivalence predicate c2 if there exist values x and y such that (and (c1 x y) (not (c2 x y))). The function hash must be acceptable for equal?, so if you use coarser equivalence than equal? (other than string-ci=?), you must always provide the function hash yourself. The hash function is always called with an appropriate bound parameter.

sizehint, when given, can be used to directly make the hash table of good size for sizehint associations, but may be ignored.

Procedure: make-symbol-hash-table [ sizehint ] → hash-table
Procedure: make-string-hash-table [ sizehint ] → hash-table
Procedure: make-string-ci-hash-table [ sizehint ] → hash-table
Procedure: make-integer-hash-table [ sizehint ] → hash-table

These procedures are provided as shorthands for creating hash tables whose keys are symbols, strings, case-insensitive strings, and integers, respectively.

Procedure: hash-table? objboolean

A predicate to test whether a given object obj is a hash table. The hash table type should be disjoint from all other types, if possible.

Procedure: alist->hash-table alist [ equal? [ hash [ sizehint ]]] → hash-table

Takes an association list alist and creates a hash table hash-table which maps the car of every element in alist to the cdr of corresponding elements in alist. equal?, hash, and sizehint are interpreted as in make-hash-table. If some key occurs multiple times in alist, it is unspecified which of the corresponding values will end up in hash-table. (Note: the choice of using cdr (instead of cadr) for values tries to strike balance between the two approaches: using cadr would render this procedure unusable for cdr alists, but not vice versa.)

Reflective queries

Procedure: hash-table-equivalence-function hash-table

Returns the equivalence predicate used for keys of hash-table.

Procedure: hash-table-hash-function hash-table

Returns the hash function used for keys of hash-table.

Dealing with single elements

Procedure: hash-table-ref hash-table key [ thunk ] → value

This procedure returns the value associated to key in hash-table. If no value is associated to key and thunk is given, it is called and its value is returned; if thunk is not given, an error is signalled. Given a good hash function, this operation should have an (amortised) complexity of O(1) with respect to the number of associations in hash-table. (Note: this rules out implementation by association lists or fixed-length hash tables.)

Procedure: hash-table-ref/default hash-table key defaultvalue

Equivalent to (hash-table-ref hash-table key (lambda () default)).

Procedure: hash-table-set! hash-table key value → undefined

This procedure sets the value associated to key in hash-table. The previous association (if any) is removed. Given a good hash function, this operation should have an (amortised) complexity of O(1) with respect to the number of associations in hash-table. (Note: this rules out implementation by association lists or fixed-length hash tables.)

Procedure: hash-table-delete! hash-table key → undefined

This procedure removes any association to key in hash-table. It is not an error if no association for that key exists; in this case, nothing is done. Given a good hash function, this operation should have an (amortised) complexity of O(1) with respect to the number of associations in hash-table. (Note: this rules out implementation by association lists or fixed-length hash tables.)

Procedure: hash-table-exists? hash-table keyboolean

This predicate tells whether there is any association to key in hash-table. Given a good hash function, this operation should have an (amortised) complexity of O(1) with respect to the number of associations in hash-table. (Note: this rules out implementation by association lists or fixed-length hash tables.)

Procedure: hash-table-update! hash-table key function [ thunk ] → undefined

Semantically equivalent to, but may be implemented more efficiently than, the following code:

(hash-table-set! hash-table key
                 (function (hash-table-ref hash-table key thunk)))

Procedure: hash-table-update!/default hash-table key function default → undefined

Equivalent to (hash-table-update! hash-table key function (lambda () default)).

Dealing with the whole contents

Procedure: hash-table-size hash-tableinteger

Returns the number of associations in hash-table. This operation must have a complexity of O(1) with respect to the number of associations in hash-table.

Procedure: hash-table-keys hash-tablelist

Returns a list of keys in hash-table. The order of the keys is unspecified.

Procedure: hash-table-values hash-tablelist

Returns a list of values in hash-table. The order of the values is unspecified, and is not guaranteed to match the order of keys in the result of hash-table-keys.

Procedure: hash-table-walk hash-table proc → unspecified

proc should be a function taking two arguments, a key and a value. This procedure calls proc for each association in hash-table, giving the key of the association as key and the value of the association as value. The results of proc are discarded. The order in which proc is called for the different associations is unspecified.

(Note: in some implementations, there is a procedure called hash-table-map which does the same as this procedure. However, in other implementations, hash-table-map does something else. In no implementation that I know of, hash-table-map does a real functorial map that lifts an ordinary function to the domain of hash tables. Because of these reasons, hash-table-map is left outside this SRFI.)

Procedure: hash-table-fold hash-table f init-valuefinal-value

This procedure calls f for every association in hash-table with three arguments: the key of the association key, the value of the association value, and an accumulated value, val. val is init-value for the first invocation of f, and for subsequent invocations of f, the return value of the previous invocation of f. The value final-value returned by hash-table-fold is the return value of the last invocation of f. The order in which f is called for different associations is unspecified.

Procedure: hash-table->alist hash-tablealist

Returns an association list such that the car of each element in alist is a key in hash-table and the corresponding cdr of each element in alist is the value associated to the key in hash-table. The order of the elements is unspecified.

The following should always produce a hash table with the same mappings as a hash table h:

(alist->hash-table (hash-table->alist h)
                        (hash-table-equivalence-function h)
                        (hash-table-hash-function h))

Procedure: hash-table-copy hash-tablehash-table

Returns a new hash table with the same equivalence predicate, hash function and mappings as in hash-table.

Procedure: hash-table-merge! hash-table1 hash-table2hash-table

Adds all mappings in hash-table2 into hash-table1 and returns the resulting hash table. This function may modify hash-table1 destructively.

Hashing

Hashing means the act of taking some value and producing a number from the value. A hash function is a function that does this. Every equivalence predicate e has a set of acceptable hash functions for that predicate; a hash funtion hash is acceptable iff (e obj1 obj2)(= (hash obj1) (hash obj2)).

A hash function h is good for a equivalence predicate e if it distributes the result numbers (hash values) for non-equal objects (by e) as uniformly as possible over the numeric range of hash values, especially in the case when some (non-equal) objects resemble each other by e.g. having common subsequences. This definition is vague but should be enough to assert that e.g. a constant function is not a good hash function.

When the definition of make-hash-table above talks about an appropriate hashing function for e, it means a hashing function that gives decent performance (for the hashing operation) while being both acceptable and good for e. This definition, too, is intentionally vague.

Procedure: hash object [ bound ] → integer

Produces a hash value for object in the range ( 0, bound (. If bound is not given, the implementation is free to choose any bound, given that the default bound is greater than the size of any imaginable hash table in a normal application. (This is so that the implementation may choose some very big value in fixnum range for the default bound.) This hash function is acceptable for equal?.

Procedure: string-hash string [ bound ] → integer

The same as hash, except that the argument string must be a string.

Procedure: string-ci-hash string [ bound ] → integer

The same as string-hash, except that the case of characters in string does not affect the hash value produced.

Procedure: symbol-hash symbol [ bound ] → integer

The same as hash, except that the argument symbol must be a symbol. Implementations are encouraged to provide this function as a builtin.

Procedure: hash-by-identity object [ bound ] → integer

The same as hash, except that this function is only guaranteed to be acceptable for eq?. The reason for providing this function is that it might be implemented significantly more efficiently than hash. Implementations are encouraged to provide this function as a builtin.

Implementation

This implementation relies on SRFI-9 for distinctness of the hash table type, and on SRFI-23 for error reporting. Otherwise, the implementation is pure R5RS.


(define *default-bound* (- (expt 2 29) 3))

(define (%string-hash s ch-conv bound)
  (let ((hash 31)
	(len (string-length s)))
    (do ((index 0 (+ index 1)))
      ((>= index len) (modulo hash bound))
      (set! hash (modulo (+ (* 37 hash)
			    (char->integer (ch-conv (string-ref s index))))
			 *default-bound*)))))

(define (string-hash s . maybe-bound)
  (let ((bound (if (null? maybe-bound) *default-bound* (car maybe-bound))))
    (%string-hash s (lambda (x) x) bound)))

(define (string-ci-hash s . maybe-bound)
  (let ((bound (if (null? maybe-bound) *default-bound* (car maybe-bound))))
    (%string-hash s char-downcase bound)))

(define (symbol-hash s . maybe-bound)
  (let ((bound (if (null? maybe-bound) *default-bound* (car maybe-bound))))
    (%string-hash (symbol->string s) (lambda (x) x) bound)))

(define (hash obj . maybe-bound)
  (let ((bound (if (null? maybe-bound) *default-bound* (car maybe-bound))))
    (cond ((integer? obj) (modulo obj bound))
	  ((string? obj) (string-hash obj bound))
	  ((symbol? obj) (symbol-hash obj bound))
	  ((real? obj) (modulo (+ (numerator obj) (denominator obj)) bound))
	  ((number? obj)
	   (modulo (+ (hash (real-part obj)) (* 3 (hash (imag-part obj))))
		   bound))
	  ((char? obj) (modulo (char->integer obj) bound))
	  ((vector? obj) (vector-hash obj bound))
	  ((pair? obj) (modulo (+ (hash (car obj)) (* 3 (hash (cdr obj))))
			       bound))
	  ((null? obj) 0)
	  ((not obj) 0)
	  ((procedure? obj) (error "hash: procedures cannot be hashed" obj))
	  (else 1))))

(define hash-by-identity hash)

(define (vector-hash v bound)
  (let ((hashvalue 571)
	(len (vector-length v)))
    (do ((index 0 (+ index 1)))
      ((>= index len) (modulo hashvalue bound))
      (set! hashvalue (modulo (+ (* 257 hashvalue) (hash (vector-ref v index)))
			      *default-bound*)))))

(define %make-hash-node cons)
(define %hash-node-set-value! set-cdr!)
(define %hash-node-key car)
(define %hash-node-value cdr)

(define-record-type <srfi-hash-table>
  (%make-hash-table size hash compare associate entries)
  hash-table?
  (size hash-table-size hash-table-set-size!)
  (hash hash-table-hash-function)
  (compare hash-table-equivalence-function)
  (associate hash-table-association-function)
  (entries hash-table-entries hash-table-set-entries!))

(define *default-table-size* 64)

(define (make-hash-table . args)
  (let* ((comparison
	   (if (null? args) equal? (car args)))
	 (hash
	   (or (and (not (null? args))
		    (not (null? (cdr args)))
		    (cadr args))
	       (and (eq? comparison string=?) string-hash)
	       (and (eq? comparison string-ci=?) string-ci-hash)
	       hash))
	 (size
	   (if (or (null? args) (null? (cdr args)) (null? (cddr args)))
	     *default-table-size* (caddr args)))
	 (association
	   (or (and (eq? comparison eq?) assq)
	       (and (eq? comparison eqv?) assv)
	       (and (eq? comparison equal?) assoc)
	       (letrec
		 ((associate
		    (lambda (val alist)
		      (cond ((null? alist) #f)
			    ((comparison val (caar alist)) (car alist))
			    (else (associate val (cdr alist)))))))
		 associate))))
    (%make-hash-table 0 hash comparison association (make-vector size '()))))

(define (make-hash-table-maker comp hash)
  (lambda args (apply make-hash-table (cons comp (cons hash args)))))
(define make-symbol-hash-table
  (make-hash-table-maker eq? symbol-hash))
(define make-string-hash-table
  (make-hash-table-maker string=? string-hash))
(define make-string-ci-hash-table
  (make-hash-table-maker string-ci=? string-ci-hash))
(define make-integer-hash-table
  (make-hash-table-maker = modulo))

(define (%hash-table-hash hash-table key)
  ((hash-table-hash-function hash-table)
     key (vector-length (hash-table-entries hash-table))))

(define (%hash-table-find entries associate hash key)
  (associate key (vector-ref entries hash)))

(define (%hash-table-add! entries hash key value)
  (vector-set! entries hash
	       (cons (%make-hash-node key value)
		     (vector-ref entries hash))))

(define (%hash-table-delete! entries compare hash key)
  (let ((entrylist (vector-ref entries hash)))
    (cond ((null? entrylist) #f)
	  ((compare key (caar entrylist))
	   (vector-set! entries hash (cdr entrylist)) #t)
	  (else
	    (let loop ((current (cdr entrylist)) (previous entrylist))
	      (cond ((null? current) #f)
		    ((compare key (caar current))
		     (set-cdr! previous (cdr current)) #t)
		    (else (loop (cdr current) current))))))))

(define (%hash-table-walk proc entries)
  (do ((index (- (vector-length entries) 1) (- index 1)))
    ((< index 0)) (for-each proc (vector-ref entries index))))

(define (%hash-table-maybe-resize! hash-table)
  (let* ((old-entries (hash-table-entries hash-table))
	 (hash-length (vector-length old-entries)))
    (if (> (hash-table-size hash-table) hash-length)
      (let* ((new-length (* 2 hash-length))
	     (new-entries (make-vector new-length '()))
	     (hash (hash-table-hash-function hash-table)))
	(%hash-table-walk
	  (lambda (node)
	    (%hash-table-add! new-entries
			      (hash (%hash-node-key node) new-length)
			      (%hash-node-key node) (%hash-node-value node)))
	  old-entries)
	(hash-table-set-entries! hash-table new-entries)))))

(define (hash-table-ref hash-table key . maybe-default)
  (cond ((%hash-table-find (hash-table-entries hash-table)
			   (hash-table-association-function hash-table)
			   (%hash-table-hash hash-table key) key)
	 => %hash-node-value)
	((null? maybe-default)
	 (error "hash-table-ref: no value associated with" key))
	(else ((car maybe-default)))))

(define (hash-table-ref/default hash-table key default)
  (hash-table-ref hash-table key (lambda () default)))

(define (hash-table-set! hash-table key value)
  (let ((hash (%hash-table-hash hash-table key))
	(entries (hash-table-entries hash-table)))
    (cond ((%hash-table-find entries
			     (hash-table-association-function hash-table)
			     hash key)
	   => (lambda (node) (%hash-node-set-value! node value)))
	  (else (%hash-table-add! entries hash key value)
		(hash-table-set-size! hash-table
				       (+ 1 (hash-table-size hash-table)))
		(%hash-table-maybe-resize! hash-table)))))

(define (hash-table-update! hash-table key function . maybe-default)
  (let ((hash (%hash-table-hash hash-table key))
	(entries (hash-table-entries hash-table)))
    (cond ((%hash-table-find entries
			     (hash-table-association-function hash-table)
			     hash key)
	   => (lambda (node)
	        (%hash-node-set-value! node (function (%hash-node-value node)))))
	  ((null? maybe-default)
	   (error "hash-table-update!: no value exists for key" key))
	  (else (%hash-table-add! entries hash key
				  (function ((car maybe-default))))
		(hash-table-set-size! hash-table
				       (+ 1 (hash-table-size hash-table)))
		(%hash-table-maybe-resize! hash-table)))))

(define (hash-table-update!/default hash-table key function default)
  (hash-table-update! hash-table key function (lambda () default)))

(define (hash-table-delete! hash-table key)
  (if (%hash-table-delete! (hash-table-entries hash-table)
			   (hash-table-equivalence-function hash-table)
			   (%hash-table-hash hash-table key) key)
    (hash-table-set-size! hash-table (- (hash-table-size hash-table) 1))))

(define (hash-table-exists? hash-table key)
  (and (%hash-table-find (hash-table-entries hash-table)
			 (hash-table-association-function hash-table)
			 (%hash-table-hash hash-table key) key) #t))

(define (hash-table-walk hash-table proc)
  (%hash-table-walk
    (lambda (node) (proc (%hash-node-key node) (%hash-node-value node)))
    (hash-table-entries hash-table)))

(define (hash-table-fold hash-table f acc)
  (hash-table-walk hash-table 
		       (lambda (key value) (set! acc (f key value acc))))
  acc)

(define (alist->hash-table alist . args)
  (let ((hash-table
	  (make-hash-table
	    (and (not (null? args)) (car args))
	    (and (not (null? args)) (not (null? (cdr args))) (cadr args))
	    (if (and (not (null? args)) (not (null? (cdr args))) (not (null? (cddr args))))
	      (caddr args)
	      (max *default-table-size* (* 2 (length alist)))))))
    (for-each (lambda (elem) (hash-table-set! hash-table (car elem) (cdr elem)))
	      alist)
    hash-table))

(define (hash-table->alist hash-table)
  (hash-table-fold hash-table
		   (lambda (key val acc) (cons (cons key val) acc)) '()))

(define (hash-table-copy hash-table)
  (let ((new (make-hash-table (hash-table-equivalence-function hash-table)
  			      (hash-table-hash-function hash-table)
			      (* 2 (hash-table-size hash-table)))))
    (hash-table-walk hash-table (lambda (key value) (hash-table-set! new key value)))
    new))

(define (hash-table-merge! hash-table1 hash-table2)
  (hash-table-walk
    hash-table2
    (lambda (key value) (hash-table-set! hash-table1 key value)))
  hash-table1)

(define (hash-table-keys hash-table)
  (hash-table-fold hash-table (lambda (key val acc) (cons key acc)) '()))

(define (hash-table-values hash-table)
  (hash-table-fold hash-table (lambda (key val acc) (cons val acc)) '()))


Copyright

Copyright © Panu Kalliokoski (2005). All Rights Reserved.

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the Software), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED AS IS, WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.


Editor: David Van Horn
Last modified: Mon Apr 25 12:00:26 EDT 2005