doc.txt

Unlib : A Utility Library by Untyped

_Unlib_ : A Utility Library by Untyped
======================================

By  Dave Gurnell (djg at untyped dot com)
and Noel Welsh   (noel at untyped dot com)

This manual documents Unlib version 2.0
Time-stamp: <2008-03-28 17:00:57 noel>

Keywords: _unlib_




==============================================================
THIS DOCUMENT HAS BEEN SUPERSEDED BY SCRIBBLE DOCUMENTATION,
AND IS NOW PARTIALLY OUT OF DATE. THIS DOCUMENT WILL BE DELETED
IN THE FUTURE.
==============================================================





Introduction
============

Unlib is a library of utility functions that have arisen
from work within Untyped (www.untyped.com).  It is available
under the LGPL (same licence as PLT Scheme).  In addition to
releases via PLaneT the source may be checked out of
Subversion.  The Subversion URL is:

  http://svn.untyped.com/unlib/



Unlib API
=========

base.ss
-------

> exn:unlib

Basic exception thrown when Unlib encounters an error

> exn:fail:unlib

Basic exception thrown when Unlib encounters an
unrecoverable error


contract.ss
-----------

> symbol-or-false?
> string-or-false?
> number-or-false?
> integer-or-false?
 : any -> boolean

These functions are useful in constructing contracts. They all return #t
if the input value matches those given in their name, and return #f
otherwise.

E.g.

  (string-or-false "foo") -> #t
  (string-or-false 2) -> #f


> arity/c :: (arity/c number)
arity/c : natural-number -> (any -> boolean)

arity/c creates a contract that checks for a procedure that accepts the 
given number of arguments.  The number of arguments is checked using
procedure-arity-includes?



time.ss
-------

> time-tai?
time-tai? : any -> boolean

True if the value is a SRFI-19 time-tai value


> time-utc?
time-utc? : any -> boolean

True if the value is a SRFi-19 time-utc value


> time-duration?
time-duration? : any -> boolean

True if the value is a SRFI-19 time-duration value


> (leap-year? year)
leap-year? : number -> (U #t #f)

True if the number represents a leap year, false otherwise

Example

  (leap-year? 2000) -> #t


> (days-in-month month [year])
days-in-month : number [number] -> number

Returns the number of days in the given month and year.  The
year defaults to 2001 (a non-leap year) if not specified.  Raises
exn:fail:unlib if the month is not in the range 1-12.

Example

  (days-in-month 2 2000) -> 29


> date-valid?
date-valid? : srfi-19-date -> boolean

True if the given SRFI-19 date is a valid date (i.e. the day, month,
year, and so forth are in the valid ranges), and false otherwise.


> time-weekday?
time-weekday? : (U time-tai time-utc) -> boolean

True if the given time represents a weekday date (that is, Monday to
Friday), and false otherwise.


> time->date
time->date : (U time-tai time-utc) -> date

Converts the given time to a date.


> seconds->ago-string :: (seconds->ago-string then [now])
seconds->ago-string : integer [integer] -> string

Takes an integer seconds value (like the value returned by current-seconds) and,
optionally, a second argument of the same type representing the current seconds,
and returns a string like:
  
  n second(s) ago
  n minute(s) ago
  n hour(s) ago
  n day(s) ago

Raises exn:fail:unlib if now is before then.


> time->ago-string : (time-tai->ago-string then [now]
time->ago-string : (U time-tai time-utc) [(U time-tai time-utc)] -> string

Takes a time-tai or time-utc (and, optionally, another argument of the same type representing
the current time) and returns a string like that of seconds->ago-string.


> current-time-zone-offset 
current-time-zone-offset : -> integer

Returns the current time-zone offset in seconds.


> current-year
current-year : -> integer

Returns the current year.



debug.ss
--------

> debug-log
debug-log : log

The log (see log.ss) used by the debug functions.


> debug-enabled?
parameter debug-enabled? : boolean

Parameter that turns debugging logging on and off.


> (debug message value)
debug : string 'a -> 'a

Prints a debug message consisting of message and value, and
returns value.  You can wrap it around an expression to
debug it without affecting the semantics of the code.

Example

   (debug "Message" (+ 1 2 3))


> (let-debug ((var val) ...) expr ...)
Syntax

Evaluates as a let, printing the values of var

Example

  (let-debug ([a 1]
              [b 2])
    (+ a b))


> (let*-debug ((var val) ...) expr ...)
Syntax

Evaluates as a let*, printing the values of var

Example

  (let*-debug ([a 1]
               [b (+ a 1)])
    (+ a b))


> (letrec-debug ((var val) ...) expr ...)
Syntax

Evaluates as a letrec, printing the values of var

Example

  (letrec-debug ([odd?
                  (lambda (n)
                    (if (zero? n)
                        #t
                        (even? (sub1 n))))]
                 [even?
                  (lambda (n)
                    (if (zero? n)
                        #f
                        (odd? (sub1 n))))])
    (odd? 7))


> (define-debug var val)
Syntax

Evaluates as a define, printing the value of var

Example

  (define-debug a 2)



exn.ss
------

> raise-exn :: (raise-exn exn message arg1 arg2 ...)
Syntax

Raises the given exception with the given message.  Handles
the conversion of message to an immutable string and
capturing continuation marks. Arguments after the message
are passed to the constructor of the exception.

Example:

  (raise-exn exn:unlib "Something wicked this way comes")


> reraise-exn :: (reraise-exn old-exn new-exn new-message other ...)
syntax reraise-exn : exn exn string message 'a ...

Raises new-exn with new-message appended to old exn's message (with a 
seperating ": ") and old-exn's continuation marks.  The additional
arguments, if any, are passed to the constructor of new-exn.

Example

  (with-handlers
    [(exn:fail?
     (lambda (e) (reraise-exn e exn:fail:unlib "This went wrong"))]
    ...)


> display-exn
display-exn : exn -> void

Displays the exn in a human readable form to the current output port.

file.ss
-------

> (make-directory-tree tree)
make-directory-tree : (tree = (U (list-of tree) string)) -> void

Creates a directory tree in the current directory that
matches tree.  Existing directories are ignored.

Example

  (make-directory-tree '("a" ("b" "c" ("d"))))

  Creates the directory tree:
    a / b
      / c / d
    

> (make-non-conflicting-filename path filename)
make-non-conflicting-filename : path string -> string

Returns a string representing a filename that is guaranteed
to name a file that does currently not exist in path.  If
filename doesn't exist in path, filename is returned,
otherwise digits are appended to the stem of filename till a
unique name is generated.

Example

   my-file.txt becomes:
  
    my-file.txt if my-file.txt doesn't exist in path
    my-file1.txt if my-file.txt exists in path
    my-file2.txt if my-file.txt and my-file1.txt exists in path


> (make-non-conflicting-path path filename)
make-non-conflicting-path : path string -> path

As make-non-conflicting-filename but returns

  (build-path path (make-non-conflicting-filename path filename))


> read-file->string
read-file->string : (U path string) -> string

Reads the contents of the file to a string.  See the port.plt collection
on PLaneT for more advanced functions along these lines.


> concatenate-files :: (concatenate-files dest src)
concactenate-files : (U string path) (list-of (U string path)) -> void

Concatenates (appends) the contents of all the files named in src to the
file named in dest.



generator.ss
------------

There is no doubt that lists are useful structures for representing
many kinds of data, and that folds and maps are a quick, convenient
way of performing arbitrary bits of list manipulation.

The main problem with the list/fold/map approach is the number of
temporary lists generated in the process, which can take up a large
amount of memory.

Generators are a half-way-house between lists and streams that aim
to reduce memory overhead when large data sources are involved.

A generator is a stream-like accessor that can be repeatedly called
to return new values from its source. A special "generator-end" value
is returned to indicate that the source has been exhausted.

For convenience we write a generator of a type "a" as follows:

    (gen-> a) === (-> (U a generator-end))

This library provides convenient ways of:

  - producing generators from lists
  - combining generators to form other generators 
    (c.f. fold, map and so on)
  - accumulating results from generators
    (e.g. back into lists)


> generator-end 
generator-end : symbol

A unique symbol that indicates a generator has finished producing values


> gen-> :: (gen-> value-contract)
syntax gen->

Syntax that expands into a contract for a generator that produces 
value-contract or generator-end


> generator-end? 
generator-end? : any -> boolean

True if the value is the generator-end symbol


> generator-map :: (generator-map fn gen0 gen1 ...)
generator-map : ('a 'b ... -> 'c) (gen-> 'a) (gen-> 'b) ... -> (gen-> 'c)

The generator equivalent of map.  Creates a new generator that returns 
fn applied to the values from gen0 etc.  The created generator ends as 
as soon as any of the source generators end.


> generator-fold-map :: (generator-fold-map fn seed gen0 gen1 ...)
generator-fold-map : ('a 'b ... 'k -> 'k) 'k (gen-> 'a) (gen-> 'b) ... 
                  -> (gen-> 'k)

A generator equivalent of fold.  The returned generator returns the
values of seed for each application of fn, stopping when any of the 
source generators stop.


> generator-filter :: (generator-filter pred src)
generator-filter : ('a -> boolean) (gen-> 'a) -> (gen-> 'a)

Given a predicate and a source generator, creates a generator that
generates the values for which the predicate is not #f.


> generator-filter-map :: (generator-filter-map fn src)
generator-filter-map : ('a -> (U 'b #f)) (gen-> 'a) -> (gen-> 'b)

Creates a generates that generates the non-#f values of fn applied to 
the values generates by src.


> generator-remove-duplicates 
generator-remove-duplicates : (gen-> 'a) [('a 'a -> boolean)] -> (gen-> 'a)

Creates a generator that filters out values that occur more than once 
in succession.  For example, if the source generator generates the 
values

  1 2 2 3 2 2 4 4 4 generator-end

the values with duplicates removed will be

  1 2 3 2 4 generator-end

The optional second argument is a function to use for equality testing.
It defaults to equal?


> generator-debug
generator-debug : string (gen-> 'a) -> (gen-> 'a)

Creates a generator equivalent to the source, except values are 
displayed on the current output port as they are generated.


> generator-for-each :: (generator-for-each fn gen0 gen1 ...)
generator-for-each : ('a 'b ... -> void) (gen-> 'a) (gen-> 'b) ... -> void

Applies fn for side-effect to the values generated by gen0 gen1 ...
Processing stops when the first source generator stops


> generator-fold :: (generator-fold fn seed gen0 gen1 ...)
generator-fold : ('a 'b ... 'k -> 'k) 'k (gen-> 'a) (gen-> 'b) ... 
              -> 'k

Like generator-fold-map but only the final application of fn is returned.


> list->generator
list->generator : (listof 'a) -> (gen-> 'a)

Creates a generator that generates the value in the given list.


> generator->list
generator->list : (gen-> 'a) -> (listof 'a)

A convenience form of generator-fold that collects the generated
values into a list (in the order generated).


> generator-project
generator-project : (list boolean ...) (gen-> (list any ...)) [(any any -> boolean)]
                 -> (gen-> (list any ... (listof (list any ...))))

Does the equivalent of a projection (from relational algebra) on
the values returned by the generator argument, called the "target".

The list argument specifies a "mask": #t values in the mask correspond
(by position) to "key" values in lists from the target, while #f values 
correspond to "nonkey" values.

The target is polled and the members returned are partitioned into keys and
nonkeys. The target is then polled repeatedly until it returns a list where
the keys differ from those stored. At this point, the generator emits a list
of the matching keys and a list of the lists of nonkeys:

    (list key ... (listof (list nonkey ...)))

The optional third argument to the procedure allows you to specify the 
predicate used for checking key equality: "eq?" is used by default.

This procedure is useful (in conjunction with g:map and map and match-lambda)
for iterating over sets of related results returned by database queries.

See generator-test.ss for examples.


gen.ss
------

Provides abbreviated names for the generator library above.  Exported
names are:

  generator.ss                 gen.ss
  ------------------------------------------------
   list->generator             list->generator

   generator-end               g:end
   generator-end?              g:end?

   generator-map               g:map               
   generator-fold-map          g:fold-map
   generator-filter            g:filter            
   generator-filter-map        g:filter-map        
   generator-remove-duplicates g:remove-duplicates 
   generator-debug             g:debug             
   generator-fold              g:fold
   generator-for-each          g:for-each          
   generator->list             g:collect           

   generator-project           g:project


hash-table.ss
-------------

Hash table utilities for PLT v3.7x and below. This library is deprecated:
it will be removed in a future version of Unlib.

> (make-hash-table/pairs . pairs)
make-hash-table/pairs : (cons 'a 'b) ... -> (hash-of 'a 'b)

Makes a hash table from the given pairs

Example

  (make-hash-table/pairs
   '(a . 1)
   '(b . 2)
   '(c . 3))


> (hash-table-mapped? table key)
hash-table-mapped? : (hash-of 'a 'b) 'a -> (U #t #f)

Returns true is the given hash table contains a mapping for
key.


> (hash-table-get/default table key default)
hash-table-get/default : (hash-of 'a 'b) 'a 'b -> 'b

Gets the value that the hash table maps to the given key, or
default if there is no mapping


> (hash-table-accessor table)
hash-table-accessor : (hash-of 'a 'b) -> ('a -> 'b)

Given a hash table, makes a function that takes a key and
returns the values mapped to that key, or exn:fail:unlib if
there is no mapping


> (hash-table-accessor/default table default)
hash-table-accessor/default : (hash-of 'a 'b) 'b -> ('a -> 'b)

Given a hash table, makes a function that takes a key and
returns the values mapped to that key, or default if there
is no mapping


> (hash-table-put/append! table key value)
hash-table-put/append! : (hash-of 'a (list-of 'b)) 'a 'b -> void

Appends value to the value mapped to key in table, creating
a list of one element if there is no mapping.  If the mapped
value is not a list exn:fail:unlib is raised.


> (hash-table-mutator table)
hash-table-mutator : (hash-of 'a 'b) -> ('a 'b -> void)

Given a hash table, creates a procedure that maps the key to
the value.


> (hash-table-mutator/append table)
hash-table-mutator/append : (hash-of 'a 'b) -> ('a 'b ->
void)

Given a hash table, creates a procedure that appends the
value to the values mapped to key, as given by
hash-table-put/append!


> (hash-table-find table selector [default])
hash-table-find : (hash-of 'a 'b) ('a 'b -> (U 'c #f)) [(()
-> 'd)] -> (U 'c 'd)

Applies selector to every key and value in table, returning
the first non-#f value returned by selector, or the value of
default if no non-#f value is returned.  The default value
of default is #f


> (any-keys-have-values? table)
any-keys-have-values? : (hash-of 'a 'b) -> (U #t #f)

Returns to #t if every value in table is a list, otherwise
raises exn:fail:unlib


> (key-has-values? table key)
key-has-values? : (hash-of 'a 'b) 'a -> (U #t #f)

Return #t if table maps key to a non-null list.  Raises
exn:fail:unlib otherwise.


> (hash-table->string table [delimiter])
hash-table->string : (hash-of 'a 'b) [string] -> string

Returns a string representation of table, of the form
key=value separated by delimiter.  Delimiter defaults to ,



hash.ss
-------

Hash table utilities for PLT v3.9x and above.

> (make-hash/alist data)
make-hash/alist : (listof pair) -> hash

Creates a hash table from the supplied data. The hash
uses strong keys and equal? as a key comparison function.

> (make-hasheq/alist data)
make-hasheq/alist : (listof pair) -> hash

Creates a hash table from the supplied data. The hash
uses strong keys and eq? as a key comparison function.

> (make-weak-hash/alist data)
make-weak-hash/alist : (listof pair) -> hash

Creates a hash table from the supplied data. The hash
uses weak keys and equal? as a key comparison function.

> (make-weak-hasheq/alist data)
make-weak-hasheq/alist : (listof pair) -> hash

Creates a hash table from the supplied data. The hash
uses weak keys and eq? as a key comparison function.

> (hash-set? hash key)
hash-set? : hash any -> boolean

Returns #t if hash contains the supplied key, #f otherwise.

> (hash-keys hash)
hash-keys : hash -> list

Returns a randomly ordered list of the keys stored in hash.

> (hash-values hash)
hash-values : hash -> list

Returns a randomly ordered list of the values stored in hash.



list.ss
-------

type (tree-of 'a) = (U 'a (cons tree tree))

type (alist-of 'a 'b) = (list-of (cons 'a 'b))

> mutable-cons :: (mutable-cons head rest)
mutable-cons : 'a (listof 'a) -> (listof 'a)

Cons head onto rest using a mutable cons cell.  Currently a synonym for
cons.


> (list-swap lst index1 index2)
list-swap : list natural natural -> list

Non-destructively swaps the elements at index1 and index2.  Raises
exn:fail:unlib if the indices are out of bounds or the same.


> (list-delimit list delimiter)
list-delimit : (list-of 'a) 'b -> (list-of (U 'a 'b))

Inserts delimiter between each element of list

Example:

  (list-delimit '("a" "b" "c") " ")
  -> '("a" " " "b" " " "c")


> (merge-sorted-lists list1 list2 same? less-than?)
merge-sorted-lists : (listof 'a) (listof 'a)
                     ('a 'a -> boolean) ('a 'a -> boolean)
                  -> (listof 'a)

Merges list1 and list2 in O(n) time assuming list1 and list2 are sorted
in ascending order as given by less-than?  Duplicates are detected using
same? and the item from list1 is taken when a duplicate is found.


> (char-iota count [start])
char-iota : number char

Generates count characters starting at start

Example 

  (char-iota 10 #\0)
  ->(list #\0 #\1 #\2 #\3 #\4 #\5 #\6 #\7 #\8 #\9)


> (list-pad list length [item])
list-pad : (listof 'a) natural ['a] -> (listof 'a)

Adds (via cons) copies of item to the front of list till length is
reached.  If the list is longer or the same length as that requested it
is returned.

If no item is specified #f is used.

Example:

  (list-pad (list 'a 'b 'c) 5 'z) -> (list 'z 'z 'a 'b 'c)


> (list-pad-right list length [item])
list-pad-right : (listof 'a) natural ['a] -> (listof 'a)

As list-pad above, except items are added to the end of list


> (tree-map fn tree)
tree-map : ('a -> 'b)  (tree-of 'a) -> (tree-of 'b)

Returns the tree that is the result of applying fn to each
leaf of tree


> (tree-for-each fn tree)
tree-for-each : ('a -> 'b) (tree-of 'a) -> void

Applies fn to each leaf of tree


> (assoc-value key alist)
assoc-value : 'a (alist-of 'a 'b) -> 'b

Return the value in the alist mapped to key, raising
exn:fail:unlib if no mapping is found


> (assoc-value/default key alist default)
assoc-value/default : 'a (alist-of 'a 'b) 'b -> 'b

Return the value in the alist mapped to key, of default if
no mapping is found


> (alist-accessor alist)
Syntax

Expands to a function of a key that calls assoc-value on alist


> (alist-accessor/default alist)
Syntax

Expands to a function of a key that calls assoc-value/default on alist


> (alist-set key value alist)
alist-set : 'a 'b (alist-of 'a 'b) -> (alist-of 'a 'b)

Returns an alist with the value mapped to key replaced with
value, if one exists, otherwise adding a new mapping from
key to value at the end of the list.


> (alist-mutator alist)
Syntax

Expands to a function that accepts a key and a value and
destructively the modifies alist to containing a mapping
from key to value

Example

  (define alist null)
  (define mutate! (alist-mutator alist))
  alist ;; null
  (mutate! 'a 1)
  alist :: '((a . 1))


> (alist-mutator/append alist)
Syntax

As alist-mutator, except values are appended to the end of existing
values


> (alist-map fn alist)
alist-map : ('a -> 'c) (alist-of 'a 'b) -> (alist-of 'a 'c)

Applies fn to every key and value in alist, returning a list
of the results


> (alist-for-each fn alist)
alist-for-each : ('a -> 'c) (alist-of 'a 'b) -> void

Applies fn to every key and value in alist


> (alist-merge alist1 alist2 [preference])
alist-merge : (alist-of 'a) (alist-of 'a) [(U 'first 'second)]
           -> (alist-of 'a)

Merges the two alists together.  The value of preference determines which
alist the value is taken from if the keys collide.  The default is to
take the value from the first list.


> alist-delete
alist-delete : 'a (alist-of 'a)) -> (alist-of 'a)

The SRFI-1 alist-delete, provided here for convenience



log.ss
------

> (define-struct log-stream (name) #f)
struct log-stream : symbol

A structure representing a particular stream of log messages.


> message-log : log-stream
> warning-log : log-stream
> error-log   : log-stream

The default log streams for messages, warnings, and logs, respectively.


> current-log-preamble
parameter current-log-preamble :  -> (list-of any)

Parameter that stores a thunk returning a list of values to
be included at the begining of each log message.


> current-log-port
parameter current-log-port : (U (parameter-of output-port) output-port))

Parameter that stores either a paramter storing an output-port (such as
current-output-port or current-error-port) or just an output-port.  
Defaults to current-output-port


> (with-log-preamble thunk expr ...)
Syntax

Evaluates expr ... in a dynamic environment with current-log-preamble
parameterised to thunk.


> (with-log-port port expr ...)
Syntax

Evaluates expr ... in a dynamic environment with current-log-port
parameterised to port.


> (log-message msg ...)
log-message : any ... -> integer

Logs msg and returns a time-tai as a unique identifier


> (log-warning msg ...)
log-warning : any ... -> integer

Logs msg as a warning and returns a time-tai as a unique identifier


> (log-error msg ...)
log-error : any ... -> integer

Logs msg as a error and returns a time-tai as a unique identifier


> (log-generic msg-type msg)
log-generic symbol (list-of any) -> integer

Logs msg as type msg-type, returning a time-tai as a unique identifier



number.ss
---------

> (number->symbol number)
number->symbol : number -> symbol

Converts number to a symbol.



parameter.ss
------------

> (make-guard predicate message)
make-guard : ('a -> (U #t #f)) string -> ('a -> 'a)

Make a guard for a paramter.  Given a predicate and a string
documenting the expected type, returns a function that
raises exn:fail:unlib if its argument does not match the
predicate, and otherwise returns the argument


> (define-parameter name initial-value guard with-form)
Syntax

Expands to two definitions. The first binds name to a
parameter with initial value of initial-value, and guard of
guard.  The second binds with-form to syntax (with-form
new-value expr ...) that expands to a parameterisation of
name with new-value in the dynamic environment of
expression.

Example

  (define-parameter foo 
    1 
    (make-guard (lambda (x) (or (integer? x) (not x)))
                "(U integer #f)")
    with-foo)

  Expands to

  (define foo
    (make-parameter 1 
      (make-guard (lambda (x) (or (integer? x) (not x)))
                  "(U integer #f)")))

  (define-syntax (with-foo stx)
    (syntax-case stx ()
      [(with-foo new-value exp (... ...))
       (syntax
         (parameterize ([foo new-value])
           exp (... ...)))]))

  And we can use as such:

  (with-foo 10 (foo))

  Evaluates to 10

  (with-foo "bar" 10)

  Raises exn:fail:unlib



pipeline.ss
---------

A "pipeline" allows a programmer to wrap a procedure in one
or more pieces of useful functionality. Pipelines are lists
of "stages", each of which performs some function and calls
the next stage. The last stage calls the target procedure.

An example of this (and the original reason for creating
pipelines) is request processing in a web server. The server
may consist of a number of controller procedures, each of
which serves a different page. All of these procedures may
have one or more bits of functionality in common.  For
example:

  - set up cookies
  - identify the user's browser
  - check the user's security privileges

Note that, while many of these functions will be common
across many controllers, there will be occasions where one
controller will need to do things differently from the
others.

The items above can be implemented as stages in a request
processing pipeline. A standard pipeline can be offered
site-wide, and controllers can choose to customise it where
appropriate by adding, removing or changing stages.

Stages are named so they can be uniquely referred to when
manipulating pipelines in this way. This has the added
advantage that single stages can be extracted and run out of
context with the rest of the pipeline.

More formally, given a target procedure:

     target : any ... -> any

a pipeline is a list of stages:

     pipeline : (list-of stage)

where a stage is a name and a body procedure:

     struct stage : symbol ((any ... -> any) any ... -> any)

The body procedure takes at least one argument: a
"continuation procedure" that is called to continue the
pipeline. The arguments passed to the continuation procedure
are passed on to the next stage in the pipeline. The target
procedure is considered a "pseudo stage" that is called
after all other stages.

Any stage can abort the pipeline simply by failing to call
the continuation procedure.  It is also perfectly reasonable
for stages to set up parameters, install exception handlers,
change the arguments to subsequent stages and so on.


> (call-with-pipeline pipeline procedure . args)
call-with-pipeline : pipeline (any ... -> any) any ... -> any

Calls a procedure via a pipeline. The result returned is
either the result of the procedure or that of the last stage
invoked.


> (make-stage name procedure)
struct stage : symbol ((any ... -> any) any ... -> any)

Constructs a stage with the given name and body procedure.
The first argument to the body procedure  is *always* a
continuation procedure that passes control to the next stage
in the pipeline.

The definition of stage takes advantage of MzScheme's
"structures as procedures" functionality such that stages
can be called directly as if they are procedures. For
example:

  (define-stage (my-stage continue name age)
    (printf "Hello ~a, " name)
    (continue age))

  (my-stage
   (lambda (age)
     (printf "you are ~a years old!" age))
   "Dave" 27))

would print:

     Hello Dave, you are 27 years old!


> (make-stage name procedure)
make-stage : symbol ((any ... -> any) any ... -> any) -> stage

Creates a stage with the specified name and procedure.


> (define-stage (name a b c ...) expr ...)
> (define-stage (name a b c . rest) expr ...)
Syntax

Shorthand syntax for make-stage with support for fixed argument lists and rest arguments.


> (stage? val)
stage? : any -> (U #t #f)

Returns #t if val is stage, #f otherwise.


> (stage-name stage)
stage-name : stage -> symbol

Returns the name of the stage.


> (find-stage pipeline name)
find-stage : (list-of stage) symbol -> (U stage #f)

Returns the appropriately named stage in the specified
pipeline, or #f if such a stage cannot be found.


> (replace-stage pipeline name)
replace-stage : (list-of stage) stage -> (list-of stage)

Replaces the equivalently named stage in the supplied
pipeline (if such a stage can be found).


> (delete-stage pipeline name)
delete-stage : (list-of stage) symbol -> (list-of stage)

Deletes the appropriately named stage from the supplied
pipeline (if such a stage can be found).



preprocess.ss
-------------

> (apply-template template bindings)
apply-template : (U string port) (alist-of symbol any) -> string

Applies template to the given bindings, returning a string
of the results.  Template is a port or string representing
an mzpp template (see (lib "mzpp.ss" "preprocessor")).

Examples:

  (define tmpl (open-input-string "<< (+ 1 2) >>"))
  (apply-template tmpl '()) ;; evaluates to "3\n"

  (define tmpl (open-input-string "<< dummy >>"))
  (apply-template tmpl '((dummy . 3))) ;; evaluates to "3\n"

Note that the bindings are eval'ed, so they should be
s-expressions that evaluate to the values you want in the
template.  Confused?  Here's an example:

Say you want the binding foo to be the value '(1 2 3)
If you call

  (apply-template tmpl '(foo . (1 2 3)))

you will get an error, as the s-expression '(1 2 3) evals to
an application of 1 to the arguments 2 3.  As 1 is not a
function this is an error.  What you want is

  (apply-template tmpl '(foo . (list 1 2 3)))

as the s-expression '(list 1 2 3) evaluates to the list '(1
2 3)

See the print-convert library in MzLib for a generic way of
converting values to their eval-able representation.



profile.ss
----------

> (profile message fn [arg ...])
profile: string (any ... -> any1) any ... -> any1

Applys fn to args and returns the result.  Measures the time
taken to apply fn and logs it (see log.ss above) with a
message that begins with "Profile" and then the message
passed to profile.



string.ss
---------

> (string-namecase string)
string-namecase : string -> string

Similar to string-titlecase buts deals with various special
cases common in European names


> (ensure-string string-or-bytes)
ensure-string : (U string bytes) -> string

If the input is bytes it is converted to a string using the
UTF-8 conversion.  Useful in servlets where the web server
may provide bytes or strings depending on the encoding

> (string-delimit strings delimiter [#:prefix prefix] [#:suffix suffix])
string-delimit : (listof string) string #:prefix string #:suffix string
              -> string
              
Creates a string with delimiter inserted between every element of 
strings.  Prefix and suffix, if given, are inserted at the beginning 
and end of the string respectively.

Example:

   (string-delimit (list "a" "b" "c")
                   ", "
                   #:prefix "["
                   #:suffix "]")
   ->
   "[a, b, c]"



symbol.ss
---------

> (symbol-append symbol ...)
symbol-append : symbol ... -> symbol

Appends several symbols, creating a new symbol

> (symbol-upcase symbol)
symbol-upcase : symbol -> symbol

Converts a symbol to uppercase.

> (symbol-downcase symbol)
symbol-downcase : symbol -> symbol

Converts a symbol to lowercase.


syntax.ss
---------

> (syntax-map fn stx)
syntax-map : (stx -> 'a) stx -> (list-of 'a)

Applies fn to every element in stx, returning a list of the results


> (syntax-append-map fn stx)
syntax-append-map : (stx -> (list-of 'a)) stx -> (list-of 'a)

Applies fn to every element in stx. appending the results.


> (symbolic-identifier=? stx1 stx2)
symbolic-identifier=? : stx stx -> (U #t #f)

Returns #t if the datums that stx1 and stx2 represent are eq?


> (atom->string atom)
atom->string : (U string symbol number stx) -> string

Converts atom to a string


> (make-syntax-symbol stx . args)
make-syntax-symbol : stx (U string symbol number stx) ... -> stx

Concatentates the string representations of args and
converts to syntax in the lexical context of stx.  Useful
for constructing identifiers.  For example:

  (make-syntax-symbol stx 'make- 'foo)

creates syntax of make-foo in the lexical context of stx



trace.ss
--------

> (define-traced (name arg ...) expr1 expr2 ...)
> (define-traced name (lambda (arg1 arg2 ...) expr1 expr2 ...))
> (define-traced name (opt-lambda (arg1 arg2 ...) expr1 expr2 ...))
Syntax

Define name in the usual way, except procedure entry and
exit is logged to the current-output-port.  

> (lambda-traced (arg ...) expr ...)

Like define-traced above, but expands to a lambda, not a
define, and output goes via the same system as debug.ss
above.


write-through-cache.ss
----------------------

A write-through cache is essentially a hash-table that calls user 
supplied functions to load and store data.  This is useful for, e.g.,
caching database values so that writes are always sent to the database
(so it is always up-to-date) but data is only loaded as needed.

The write-through cache associates a fixed lifespan to all
stored data, so data in the cache will eventually be
removed.  This ensures data is the cache does not get too
old, and additionally that memory consumption is constrained.

> (make-write-through-cache load store
                            [#:expire expire]
                            [#:hash-code hash-code]
			    [#:lifetime lifetime])
make-write-through-cache : ('a -> 'b)
                           ('a 'b -> void) 
                           [#:expire (cache 'a 'b -> void)]
			   [#:hash-code (U 'eq 'equal)]
			   [#:lifetime natural-number]
                        -> cache

Creates a write-through cache using the given load and store functions.
Load is given a key and must return a value.  Store is given a key and 
value.

The optional expire argument is a function that takes a
cache, a key, and a value, and is called whenever a value is
removed from the cache.  The default function does nothing
but user-supplied functions can, for example, add the data
back to the cache.

The optional hash-code argument specifies if the cache uses
eq? or equal? for determining equality in the hash table.
The default is eq.  See the hash table documentation for
more on this distinction.

The optional lifetime argument is the number of seconds that
each item is stored in the cache.  The default is 3600 (1
hour).  Items are expired from the cache by periodic sweeps
of all data, and additionally will be expired if requested
using cache-get and they have expired since the most recent
sweep.


> (cache-get cache key)
cache-get : cache 'a -> 'b

Gets a value from the cache.  If the cache contains no value for key
load is called to retrieve a value.  If the cache contains a value
that value is returned and load is NOT called.


> (cache-set! cache key value)
cache-set! : cache 'a 'b -> void

Sets a value in the cache and calls store to update whatever data source
this cache represents.


> (cache-clear! cache)
cache-clear! : cache -> void

Clears all values in the cache, but does not call load or store.  Useful
to control memory consumption or for testing.



yield.ss
--------

Implements the "yield" operator of Ruby / Python using continuations.
"Yield" allows the programmer to pause the execution of a procedure,
returning a result and continuing execution from the same point in the
next invocation of the procedure.

Supports procedures with multiple arguments and return types. For 
example:

  (define calc
    (make-yieldable
      (lambda (yield)
        (lambda (a b)
          (let-values ([(c d) (yield (values (+ a b) (- a b)))])
            (values (* c d) (/ c d)))))))
          
  (define-values (w x)
    (calc 6 2))
    
  (define-values (y z)
    (calc 8 4))
    
In the example, the first call to calc returns at the call to yield.
w and x are bound to the values 8 (6+2) and 4 (6-2). The second call to 
calc resumes from the same point, and the arguments are bound to c and 
d instead of a and b. y and z are bound to the values 32 (8*4) and 2 (8/4).

> (make-yieldable yield)
make-yieldbale : (yield-proc -> target-proc) -> target-proc

yield-proc and target-proc have symmetric contracts:

    yield-proc : a b c -> d e
    target-proc : d e -> a b c

The example above demonstrates the use of this procedure.

> (yieldable yield-identifier stmt ...)
syntax

A syntactic form of make-yieldable that allows the programmer to avoid
writing so many lambdas. For example:

  (define calc
    (yieldable yield
      (lambda (a b)
        (let-values ([(c d) (yield (values (+ a b) (- a b)))])
          (values (* c d) (/ c d))))))

Check API
=========

The Check API provides types and procedures for validating data entered 
by a user. This is work-in-progress: it is likely to be updated and 
repackaged separately at some point.

The programmer writes a suite of "check" procedures that run on data
structures and check whether or not it is valid. Checks return lists of 
annotated "check-results" that report problems to the user and can be 
mapped to appropriate places in the UI.

Checks
------

Checks are blocks of code that accept some sort of data as an input
and return a list of check-result structures:

    check : any ... -> (list-of check-result)

An "atomic check" checks a single property of the data, for example:

  - the date is [not] in the right range;
  - the username does [not] have the right format;
  - the date of the outward journey is before the date of the return
    journey;
  - and so on...

Atomic checks return a single check-result embedded in a list:

    atomic-check : any ... -> (list check-result)

Compound checks group atomic checks together into blocks that return 
arbitrary lists of check-results. The Check library provides procedures 
and macros for processing lists of check-results.

Types of check-result
---------------------

A check-result is a structure of a type in the hierarchy:

  check-result
    check-success
    check-problem
      check-warning
      check-failure
      check-error

A check-success indicates that a particular constraint on the data 
holds; a check-problem indicates that the property does not hold. There 
are three types of check-problem:

- A check-failure indicates a definite problem with the data. The
  application should not allow the data to be processed until the user
  has corrected their mistakes.

- A check-warning indicates a potential problem with the data. The
  application cannot tell whether this is a real problem, or whether
  it is an odd circumstance that the user has anticipated. The 
  application should inform the user of the warning and allow them
  to proceed if they think it is okay.

- A check-error indicates that an exception has been raised when 
  processing a check. This is a convenience for the programmer: it 
  allows bugs in one check to be reported in the UI without affecting
  the results of other checks.

The types of the check-result structures are as follows (all the usual
constructors, predicates, accessors, mutators and so on are provided
by check.ss):

> (define-struct check-result (message annotations) #f)
struct check-result : string (alist-of symbol any)

The message string should be formatted to describe the result to the
user to an appropriate level of detail. Arbitrary annotations can be
attached to the result to help deliver the message to the user in an
appropriate fashion.

> (define-struct (check-problem check-result) () #f)
> (define-struct (check-success check-result) () #f)
> (define-struct (check-failure check-problem) () #f)
> (define-struct (check-warning check-problem) () #f)

These types have the same contract as check-result.

> (define-struct (check-error   check-problem) (exn) #f)
struct check-error :  string (alist-of symbol any) exn

The check-error type has an extra field to store the exception raised.

Creating check-results
----------------------

Check results are typically created with messages, and annotated and
combined later using list-processing combinators:

> (pass)
pass : -> (list check-success)

Creates a single check-success with the message "Okay".

> (fail message)
fail : string -> (list check-failure)

Creates a single check-failure with the specified message.

> (warn message)
warn : string -> (list check-warning)

Creates a single check-warning with the specified message.

> (check-with-handlers thunk)
check-with-handlers : (-> (list-of check-result))
                   -> (list-of check-result)

Wraps a thunk with a with-handlers block that catches any exceptions
and returns a 1-length list containing a check-error.

> (check-until-problems thunk ...)
check-until-problems : (-> (list-of check-result))
                       ...
                    -> (list-of check-problem)

Runs each of the check thunks in turn until one of them returns a list 
containing one or more check-problems. Returns the results of that 
thunk.

Result list combinators
-----------------------

> (check-all results ...)
check-all : (list-of check-result) ... -> (list-of check-result)

Combines several results lists into a single list. Analogous to append 
(with a nice contract on it).

> (check-with-annotations annotations thunk)
check-with-annotations : (alist-of symbol any)
                         (list-of check-result)
                      -> (list-of check-result)

Annotates a list of results with the supplied annotations. Returns the
annotated results.

> (check-problems? results)
check-problems? : (list-of check-result) -> boolean

Returns #t if there are any check-problems in results: #f otherwise.

> (check-results->problems results)
check-results->problems : (list-of check-result)
                       -> (list-of check-problem)

Filters a list of results for problems (i.e. everything except
check-successes).

> (check-results->warnings+failures+errors results)
check-results->warnings+failures+errors
     : (list-of check-result)
    -> (list-of check-warning)
       (list-of check-failure)
       (list-of check-error)

Filters a list of results for problems and splits the problems
into three lists: warnings, failures and errors.

Annotations
-----------

> (check-result-annotation? result name)
check-result-annotation : check-result symbol -> boolean

Determines whether the result has the specified annotation.

> (check-result-annotation result name)
check-result-annotation : check-result symbol -> any | exn:fail:unlib

Returns the value mapped to the specified key in the result's 
annotations. Raises exn:fail:unlib if the annotation cannot be found.

> (check-result-annotation/default result name default)
check-result-annotation : check-result symbol -> any

Returns the value mapped to the specified key in the result's 
annotations. Returns default if the annotation cannot be found.

> (annotate-check-result result annotations)
annotate-check-result : check-result (alist-of symbol any)
                     -> check-result

Adds the specified annotations to the result. Works on a single 
check-result only: the simplest way to annotate lists of results is 
using check-with-annotations.