Work in progress, do not use
This WG2 proposal defines an interface to hash tables, which are widely recognized as a fundamental data structure for a wide variety of applications. A hash table is a data structure that:
(make-hash-table = hash . args)
Returns a newly allocated hash table whose equality procedure is =, hash function is hash (defaults to a suitable function, and args mean whatever (number is initial capacity, 'immutable, 'weakkeys, 'weak-values, all implementation-dependent). The "safe" equality procedures are eq?, eqv?, equal?, string=?, string-ci=?.
(hash-table = hash key value ...)
Ditto, but with explicitly specified key/value pairs
(hash-table? obj)
Returns #t if obj is a hash table.
(hash-table-ref/default hash-table key default)
Returns the value corresponding to key in hash-table, or default if none.
(hash-table-ref hash-table key failure success)
Looks up key in hash table. If found, passes value to success procedure (default identity) and returns its value; if not found, invoke failure procedure (default is to signal an error) and return its value
(hash-table-ref!/default hash-table key default)
Same as hash-table-ref/default, but updates hash table if needed.
(hash-table-ref! hash-table key failure success)
Same as hash-table-ref, but updates hash table if needed.
(hash-table-exists? hash-table key)
Returns #t if key exists, #f if not.
(hash-table-set! hash-table key value)
Primitive mutator.
(hash-table-set*! hash-table key value ...)
Mutate key1 to value1, key2 to value2, etc.
(hash-table-set hash-table key value)
New hash table with additional (or replaced) binding.
(hash-table-set* hash-table key value ...)
Like hash-table-set*! but returns new hash table.
(hash-table-delete! hash-table key)
Delete key if it exists.
(hash-table-delete hash-table key)
Allocate copy without key.
(hash-table-update!/default hash-table key updater default)
Looks up key in hash-table, using default if not found. Invoke updater procedure and mutate the value of the same key.
(hash-table-update! hash-table key updater failure success)
Looks up key in hash-table, passing it through success procedure if found, or result of failure procedure if not found. Pass through updater, mutate the value of the same key.
(hash-table-update/default hash-table key updater default)
Same as hash-table-update!/default but returns new hash table.
(hash-table-update hash-table key updater failure success)
Same as hash-table-update! but returns new hash table.
(hash-table-clear! hash-table)
Remove all keys from hash-table as efficiently as possible.
(hash-table-length hash-table)
Return number of keys in hash-table.
(hash-table-keys hash-table)
Return list of all keys in random order.
(hash-table-values hash-table)
Return list of all values in random order, not necessarily the same order.
(hash-table-map->list hash-table proc)
Pass each key and value as separate arguments to proc, collect results in list. There is no hash-table-map in the sense of lifting a procedure to the domain of hash tables.
(hash-table-for-each hash-table proc)
The same, but discards the results.
(hash-table-copy hash-table)
Returns a fresh copy of hash-table, same equality and hash procedures.
(hash-table->alist hash-table)
Returns an alist with the keys and values of hash-table.
(alist->hash-table alist = hash . args)
Returns a new hash-table, as if with make-hash-table, initializing it with alist.
(hash-table-push! hash-table key value)
Conses value onto the value of key, empty list if no such key.
(hash-table-pop! hash-table key)
Returns car of value corresponding to key, mutates key to cdr.
(hash-table-find hash-table proc)
Proc gets called with each key and value in random order. If it returns #f, try another key and value, otherwise return what proc returns.
(hash-table-accessor/default hash-table default)
Curries hash-table-ref/default.
(hash-table-accessor hash-table success failure)
Curries hash-table-ref.
(hash-table-mutator hash-table)
Curries hash-table-set!.
(hash-table-deleter hash-table)
Curries hash-table-delete.
(hash-table-updater/default hash-table updater default)
Curries hash-table-update!/default.
(hash-table-updater hash-table updater failure success)
Curries hash-table-update!
(hash-table-union hash-table1 hash-table2)
Adds keys and values of hash-table2 to copy of hash-table1 and returns it.
(hash-table-union! hash-table ...)
Adds keys and values of hash-table2 to hash-table1 and returns it.
(hash-table-difference hash-table1 hash-table2)
Removes keys of hash-table2 from copy of hash-table1 and returns it.
(hash-table-difference! hash-table1 hash-table2)
Removes keys of hash-table2 from hash-table1 and returns it.
(hash-table=? key= value= hash-table1 hash-table2)
Returns #t if hash-tables have the same keys (in the sense of key=) and the same corresponding values (in the sense of value=).
hash-table-entries (returns both keys and values as two values)
set!* from an alist
accept key list and value list and add them all, destructive or non-destructive, analogous to CL pairlis
hash-table-value-exists?
hash-table-unfold
count and remove/remove! by a predicate
hash-table-replace! (mutates only if found)
(hash-table-map hash-table key-mapper value-mapper colliding-value-merger)
hash-table-map!, same idea
pop! should take defaults in the usual way, since it is an accessor. Issue: should it remove empty entries altogether?
union(!) should take value-merging procedure (defaults to (lambda (x y) y)).
Have a procedure that binds the equality and hash procedures into a single procedure by case-lambda. This is used to create hash tables, saving an extra argument each time. Note that this allows the author of an equality predicate to provide hashing without cluttering up the interface, and allows the predicate to be an equal peer with built-in predicate. The only bad point here is that it makes copying messy (but it's already messy, and may have to be primitive anyway).
Having both set and set!, delete and delete!, update and update! etc. may be too confusing, though it is Schemely.
Only kept around as a source of useful text to cut and paste.
These constructors provide suitable hash functions for the equivalence function specified as part of the constructor name. This proposal does not allow (semi-)arbitrary equivalence and hash functions to be specified.
(make-eq-hash-table)
Returns a newly allocated table with no associations whose equivalence function is eq?. (SRFI-69 (make-hash-table eq? hash-by-identity); R6RS (make-eq-hashtable))
(make-equal-hash-table)
Returns a newly allocated with no associations whose equivalence function is equal?. (SRFI-69 (make-hash-table equal? hash); R6RS (make-hashtable equal? equal-hash))
(make-string-hash-table)
Returns a newly allocated table with no associations whose equivalence function is string=?. (SRFI-69 (make-hash-table string=? string-hash); R6RS (make-hashtable string=? string-hash))
(make-string-ci-hash-table)
Returns a newly allocated table with no associations whose equivalence function is string-ci=?. (SRFI-69 (make-hash-table string-ci=? string-ci-hash); R6RS (make-hashtable string-ci=? string-ci-hash))
Note that there are no hash tables whose equivalence function is eqv?, because SRFI 69 does not support them. Users will have to live with eq? or equal? hash tables as the case may be.
Issue: add (hash-table-unfold ''p f g key-seed value-seed'')? Or possibly just one seed and let f generate the two values from a single value.
(hash-table-copy hash-table)
Returns a newly allocated hash table with the same equivalence predicate and associations as hash-table. (SRFI-69 hash-table-copy; R6RS (hashtable-copy hash-table #t))
(hash-table? obj)
Returns #t if obj is a hash table. (SRFI-69 hash-table?; R6RS hashtable?)
(hash-table-contains? hash-table key)
Returns #t if there is any association to key in hash-table. Must execute in amortized O(1) time. (SRFI-69 hash-table-exists?; R6RS hashtable-contains?)
(hash-table-ref hash-table key default-proc)
Returns the value associated to key in table. If no value is associated to key, default-proc is applied to no arguments and the result is returned. Must execute in amortized O(1) time, not counting the time to call default-proc if necessary. (SRFI-69 hash-table-ref/default; R6RS hashtable-ref)
(hash-table-size table)
Returns the number of associations in hash-table. Must execute in amortized O(1) time. (SRFI-69 hash-table-size; R6RS hashtable-size)
(hash-table-set! hash-table key value)
Creates a new association in hash-table that associates key with value. The previous association (if any) is deleted. It is an error if hash-table is a string or string-ci hash table and key is not a string. Must execute in amortized O(1) time. Returns an unspecified value. (SRFI-69 hash-table-set!; R6RS hashtable-set!)
(hash-table-delete! hash-table key)
Deletes any association to key in hash-table. It is not an error if no association for that key exists. Must execute in amortized O(1) time. Returns an unspecified value. (SRFI-69 hash-table-delete!; R6RS hashtable-delete!)
(hash-table-update! hash-table key procedure default-proc ])
Semantically equivalent to, but may be implemented more efficiently than, the following code:
(hash-table-set! hash-table key (procedure (hash-table-ref hash-table key default-proc)))
Must execute in amortized O(1) time. Returns an unspecified value. (SRFI-69 hash-table-update!/default; R6RS hashtable-update!)
Issue: add hash-table-clear?
(hash-table-fold procedure init hash-table)
Calls procedure for every association in hash-table with three arguments: the key of the association, the value of the association, and an accumulated value val. Val is init for the first invocation of procedure, and for subsequent invocations of procedure, the return value of the previous invocation. The value returned by hash-table-fold is the return value of the last invocation of f. The order in which procedure is called for different associations is unspecified. (SRFI-69 hash-table-fold; not in R6RS but trivially implemented: see below.)
The following features are not part of this proposal for various reasons.
Hash-table-keys is present in both SRFI 69 and R6RS, but returns a list of keys in the first, a vector of keys in the second.
Hash-table-values returns a list of values in SRFI 69; hashtable-entries returns two values, a vector of keys and a vector of values in R6RS.
Hash-table-hash-function always returns the hash function of a hash table in SRFI 69, but returns #f in the case of eq? (and eqv?) hash tables in R6RS.
Hash-table-equivalence-function returns the equivalence function, but in R6/R7RS you can't reliably check functions for identity, so you don't know what you've got.
Hashtable-clear!, which removes all associations from a hash table.
Specifying the initial capacity of a hash table.
Immutable hash tables.
Hash tables based on eqv?.
Hash-table-walk, which is the analogue of for-each for hash tables.
Conversion from a-lists to hash tables and back.
Hash-table-merge, which destructively adds all associations in a source hash table to a destination hash table.
Default values specified in the form of a thunk to call.
The current capacity (as opposed to size) of a hash table.
Rehash size and threshold in constructor, plus accessors and mutators for them.
Hash tables based on equalp (which is not in Scheme).
With-hash-table-iterator, a hash table external iterator implemented as a local macro.
Sxhash, a stable hash function.
I have added equivalences to SRFI 69 and R6RS facilities. Systems supporting either should be able to support these hash tables trivially; it will be necessary to write hash-table-fold for R6RS systems. I have adopted SRFI 69's term hash-table rather than R6RS's hashtable because of our decision in #40. Besides, the word hashtable obviously means something that can be ... hashted.
The main annoyances of this proposal for SRFI 69 programmers will be remembering to supply a third argument to hash-table-ref, and for R6RS programmers will be remembering to insert a hyphen in hashtable. Both will have to get used to the new constructors.