more radical change: merge tutorial into readme, cleanup

This commit is contained in:
Simon Cruanes 2017-02-11 14:46:56 +01:00
parent 8d671f03a1
commit 7603a78970
2 changed files with 316 additions and 393 deletions

View file

@ -2,11 +2,23 @@
:toc: macro :toc: macro
:source-highlighter: pygments :source-highlighter: pygments
What is _containers_? (take a look at the link:TUTORIAL.adoc[tutorial]! A modular, clean and powerful extension of the OCaml standard library.
or the http://c-cube.github.io/ocaml-containers/[current documentation])
In `containers` and `containers.data`, all modules abide by Containers is an extension of OCaml's standard library (under BSD license)
_pay for what you use_: only modules that are used are linked (there are no focused on data structures, combinators and iterators, without dependencies on
cross-module dependencies). unix, str or num. Every module is independent and is prefixed with 'CC' in the
global namespace. Some modules extend the stdlib (e.g. CCList provides safe
map/fold_right/append, and additional functions on lists).
Alternatively, `open Containers` will bring enhanced versions of the standard
modules into scope.
image:https://ci.cedeela.fr/buildStatus/icon?job=containers[alt="Build Status", link="http://ci.cedeela.fr/job/containers/"]
toc::[]
== Quick Summary
Containers is:
- A usable, reasonably well-designed library that extends OCaml's standard - A usable, reasonably well-designed library that extends OCaml's standard
library (in 'src/core/', packaged under `containers` in ocamlfind. Modules library (in 'src/core/', packaged under `containers` in ocamlfind. Modules
@ -18,11 +30,9 @@ cross-module dependencies).
`Containers` (intended to be opened, replaces some stdlib modules `Containers` (intended to be opened, replaces some stdlib modules
with extended ones). with extended ones).
- Several small additional libraries that complement it: - Several small additional libraries that complement it:
* `containers.data` with additional data structures that don't have an
containers.data:: with additional data structures that don't have an
equivalent in the standard library; equivalent in the standard library;
containers.iter:: with list-like and tree-like iterators; * `containers.iter` with list-like and tree-like iterators;
- Utilities around the `unix` library in `containers.unix` (mainly to spawn - Utilities around the `unix` library in `containers.unix` (mainly to spawn
sub-processes easily and deal with resources safely) sub-processes easily and deal with resources safely)
- A lightweight S-expression printer and streaming parser in `containers.sexp` - A lightweight S-expression printer and streaming parser in `containers.sexp`
@ -33,10 +43,6 @@ cross-module dependencies).
Some of the modules have been moved to their own repository (e.g. `sequence`, Some of the modules have been moved to their own repository (e.g. `sequence`,
`gen`, `qcheck`) and are on opam for great fun and profit. `gen`, `qcheck`) and are on opam for great fun and profit.
image:https://ci.cedeela.fr/buildStatus/icon?job=containers[alt="Build Status", link="http://ci.cedeela.fr/job/containers/"]
toc::[]
== Change Log == Change Log
See link:CHANGELOG.adoc[this file]. See link:CHANGELOG.adoc[this file].
@ -80,111 +86,8 @@ This code is free, under the BSD license.
== Contents == Contents
The library contains a <<core,Core part>> that mostly extends the stdlib See http://c-cube.github.io/ocaml-containers/[the documentation]
and adds a few very common structures (heap, vector), and sub-libraries and <<tutorial,the tutorial below>> for a gentle introduction.
that deal with either more specific things, or require additional dependencies.
Some structural types are used throughout the library:
gen:: `'a gen = unit -> 'a option` is an iterator type. Many combinators
are defined in the opam library https://github.com/c-cube/gen[gen]
sequence:: `'a sequence = (unit -> 'a) -> unit` is also an iterator type.
It is easier to define on data structures than `gen`, but it a bit less
powerful. The opam library https://github.com/c-cube/sequence[sequence]
can be used to consume and produce values of this type.
error:: `'a or_error = ('a, string) result = Error of string | Ok of 'a`
using the standard `result` type, supported in `CCResult`.
klist:: `'a klist = unit -> [`Nil | `Cons of 'a * 'a klist]` is a lazy list
without memoization, used as a persistent iterator. The reference
module is `CCKList` (in `containers.iter`).
printer:: `'a printer = Format.formatter -> 'a -> unit` is a pretty-printer
to be used with the standard module `Format`. In particular, in many cases,
`"foo: %a" Foo.print foo` will type-check.
[[core]]
=== Core Modules (extension of the standard library)
the core library, `containers`, now depends on
https://github.com/mjambon/cppo[cppo] and `base-bytes` (provided
by ocamlfind).
http://c-cube.github.io/ocaml-containers/[Documentation here].
- `CCHeap`, a purely functional heap structure
- `CCVector`, a growable array (pure OCaml, no C) with mutability annotations
- `CCList`, functions on lists, including tail-recursive implementations of `map` and `append` and many other things
- `CCArray`, utilities on arrays
- `CCArray_slice`, array slices
- `CCHashtbl`, `CCMap` extensions of the standard modules `Hashtbl` and `Map`
- `CCInt`
- `CCString` (basic string operations)
- `CCPair` (cartesian products)
- `CCOpt` (options, very useful)
- `CCFun` (function combinators)
- `CCBool`
- `CCFloat`
- `CCOrd` (combinators for total orderings)
- `CCRandom` (combinators for random generators)
- `CCHash` (hashing combinators)
- `CCResult` (monadic error handling, very useful)
- `CCIO`, basic utilities for IO (channels, files)
- `CCInt64,` utils for `int64`
- `CCChar`, utils for `char`
- `CCFormat`, pretty-printing utils around `Format`
=== Containers.data
- `CCBitField`, bitfields embedded in integers
- `CCCache`, memoization caches, LRU, etc.
- `CCFlatHashtbl`, a flat (open-addressing) hashtable functorial implementation
- `CCTrie`, a prefix tree
- `CCHashTrie`, a map where keys are hashed and put in a trie by hash
- `CCMultimap` and `CCMultiset`, functors defining persistent structures
- `CCFQueue`, a purely functional double-ended queue structure
- `CCBV`, mutable bitvectors
- `CCHashSet`, mutable set
- `CCPersistentHashtbl` and `CCPersistentArray`, a semi-persistent array and hashtable
(similar to https://www.lri.fr/~filliatr/ftp/ocaml/ds/parray.ml.html[persistent arrays])
- `CCMixmap`, `CCMixtbl`, `CCMixset`, containers of universal types (heterogenous containers)
- `CCRingBuffer`, a double-ended queue on top of an array-like structure,
with batch operations
- `CCIntMap`, map specialized for integer keys based on Patricia Trees,
with fast merges
- `CCGraph`, a small collection of graph algorithms
- `CCBitField`, a type-safe implementation of bitfields that fit in `int`
- `CCWBTree`, a weight-balanced tree, implementing a map interface
- `CCRAL`, a random-access list structure, with `O(1)` cons/hd/tl and `O(ln(n))`
access to elements by their index.
- `CCImmutArray`, immutable interface to arrays
=== Containers.unix
- `CCUnix`, utils for `Unix`
=== Containers.sexp
A small S-expression library.
- `CCSexp`, a small S-expression library
=== Containers.iter
Iterators:
- `CCKList`, a persistent iterator structure (akin to a lazy list, without memoization)
- `CCKTree`, an abstract lazy tree structure
=== Thread
In the library `containers.thread`, for preemptive system threads:
- `CCFuture`, a set of tools for preemptive threading, including a thread pool,
monadic futures, and MVars (concurrent boxes)
- `CCLock`, values protected by locks
- `CCSemaphore`, a simple implementation of semaphores
- `CCThread` basic wrappers for `Thread`
[[build]]
== Documentation == Documentation
@ -198,6 +101,7 @@ by version:
- http://c-cube.github.io/ocaml-containers/0.19/[0.19] - http://c-cube.github.io/ocaml-containers/0.19/[0.19]
- http://c-cube.github.io/ocaml-containers/0.17/[0.17] - http://c-cube.github.io/ocaml-containers/0.17/[0.17]
[[build]]
== Build == Build
You will need OCaml `>=` 4.01.0. You will need OCaml `>=` 4.01.0.
@ -238,3 +142,297 @@ A few guidelines:
- add tests if possible (using `qtest`). - add tests if possible (using `qtest`).
Powered by image:http://oasis.forge.ocamlcore.org/oasis-badge.png[alt="OASIS", style="border: none;", link="http://oasis.forge.ocamlcore.org/"] Powered by image:http://oasis.forge.ocamlcore.org/oasis-badge.png[alt="OASIS", style="border: none;", link="http://oasis.forge.ocamlcore.org/"]
[[tutorial]]
== Tutorial
This tutorial contains a few examples to illustrate the features and
usage of containers. We assume containers is installed and that
the library is loaded, e.g. with:
[source,OCaml]
----
#require "containers";;
----
=== Basics
We will start with a few list helpers, then look at other parts of
the library, including printers, maps, etc.
[source,OCaml]
----
(* quick reminder of this awesome standard operator *)
# (|>) ;;
- : 'a -> ('a -> 'b) -> 'b = <fun>
# open CCList.Infix;;
# let l = 1 -- 100;;
val l : int list = [1; 2; .....]
# l
|> CCList.filter_map
(fun x-> if x mod 3=0 then Some (float x) else None)
|> CCList.take 5 ;;
- : float list = [3.; 6.; 9.; 12.; 15.]
# let l2 = l |> CCList.take_while (fun x -> x<10) ;;
val l2 : int list = [1; 2; 3; 4; 5; 6; 7; 8; 9]
(* an extension of Map.Make, compatible with Map.Make(CCInt) *)
# module IntMap = CCMap.Make(CCInt);;
(* conversions using the "sequence" type, fast iterators that are
pervasively used in containers. Combinators can be found
in the opam library "sequence". *)
# let map =
l2
|> List.map (fun x -> x, string_of_int x)
|> CCList.to_seq
|> IntMap.of_seq;;
val map : string CCIntMap.t = <abstr>
(* check the type *)
# CCList.to_seq ;;
- : 'a list -> 'a sequence = <fun>
# IntMap.of_seq ;;
- : (int * 'a) CCMap.sequence -> 'a IntMap.t = <fun>
(* we can print, too *)
# Format.printf "@[<2>map =@ @[<hov>%a@]@]@."
(IntMap.print CCFormat.int CCFormat.string_quoted)
map;;
map =
[1 --> "1", 2 --> "2", 3 --> "3", 4 --> "4", 5 --> "5", 6 --> "6",
7 --> "7", 8 --> "8", 9 --> "9"]
- : unit = ()
(* options are good *)
# IntMap.get 3 map |> CCOpt.map (fun s->s ^ s);;
- : string option = Some "33"
----
=== New types: `CCVector`, `CCHeap`, `CCError`, `CCResult`
Containers also contains (!) a few datatypes that are not from the standard
library but that are useful in a lot of situations:
CCVector::
A resizable array, with a mutability parameter. A value of type
`('a, CCVector.ro) CCVector.t` is an immutable vector of values of type `'a`,
whereas a `('a, CCVector.rw) CCVector.t` is a mutable vector that
can be modified. This way, vectors can be used in a quite functional
way, using operations such as `map` or `flat_map`, or in a more
imperative way.
CCHeap::
A priority queue (currently, leftist heaps) functorized over
a module `sig val t val leq : t -> t -> bool` that provides a type `t`
and a partial order `leq` on `t`.
CCError::
An error type for making error handling more explicit (an error monad,
really, if you're not afraid of the "M"-word). It is similar to the
more recent `CCResult`, but works with polymorphic variants for
compatibility with the numerous libraries that use the same type,
that is, `type ('a, 'b) CCError.t = [`Ok of 'a | `Error of 'b]`.
CCResult::
It uses the new `result` type from the standard library (or from
the retrocompatibility package on opam), and presents an interface
similar to `CCError`. In an indeterminate amount of time, it will
totally replace `CCError`.
Now for a few examples:
[source,OCaml]
----
(* create a new empty vector. It is mutable, for otherwise it would
not be very useful. *)
# CCVector.create;;
- : unit -> ('a, CCVector.rw) CCVector.t = <fun>
(* init, similar to Array.init, can be used to produce a
vector that is mutable OR immutable (see the 'mut parameter?) *)
# CCVector.init ;;
- : int -> (int -> 'a) -> ('a, 'mut) CCVector.t = <fun>c
(* use the infix (--) operator for creating a range. Notice
that v is a vector of integer but its mutability is not
decided yet. *)
# let v = CCVector.(1 -- 10);;
val v : (int, '_a) CCVector.t = <abstr>
# Format.printf "v = @[%a@]@." (CCVector.print CCInt.print) v;;
v = [1, 2, 3, 4, 5, 6, 7, 8, 9, 10]
(* now let's mutate v *)
# CCVector.push v 42;;
- : unit = ()
(* now v is a mutable vector *)
# v;;
- : (int, CCVector.rw) CCVector.t = <abstr>
(* functional combinators! *)
# let v2 = v
|> CCVector.map (fun x-> x+1)
|> CCVector.filter (fun x-> x mod 2=0)
|> CCVector.rev ;;
val v2 : (int, '_a) CCVector.t = <abstr>
# Format.printf "v2 = @[%a@]@." (CCVector.print CCInt.print) v2;;
v2 = [10, 8, 6, 4, 2]
(* let's transfer to a heap *)
# module IntHeap = CCHeap.Make(struct type t = int let leq = (<=) end);;
# let h = v2 |> CCVector.to_seq |> IntHeap.of_seq ;;
val h : IntHeap.t = <abstr>
(* We can print the content of h
(printing is not necessarily in order, though) *)
# Format.printf "h = [@[%a@]]@." (IntHeap.print CCInt.print) h;;
h = [2,4,6,8,10]
(* we can remove the first element, which also returns a new heap
that does not contain it — CCHeap is a functional data structure *)
# IntHeap.take h;;
- : (IntHeap.t * int) option = Some (<abstr>, 2)
# let h', x = IntHeap.take_exn h ;;
val h' : IntHeap.t = <abstr>
val x : int = 2
(* see, 2 is removed *)
# IntHeap.to_list h' ;;
- : int list = [4; 6; 8; 10]
----
=== IO helpers
The core library contains a module called `CCIO` that provides useful
functions for reading and writing files. It provides functions that
make resource handling easy, following
the pattern `with_resource : resource -> (access -> 'a) -> 'a` where
the type `access` is a temporary handle to the resource (e.g.,
imagine `resource` is a file name and `access` a file descriptor).
Calling `with_resource r f` will access `r`, give the result to `f`,
compute the result of `f` and, whether `f` succeeds or raises an
error, it will free the resource.
Consider for instance:
[source,OCaml]
----
# CCIO.with_out "/tmp/foobar"
(fun out_channel ->
CCIO.write_lines_l out_channel ["hello"; "world"]);;
- : unit = ()
----
This just opened the file '/tmp/foobar', creating it if it didn't exist,
and wrote two lines in it. We did not have to close the file descriptor
because `with_out` took care of it. By the way, the type signatures are:
[source,OCaml]
----
val with_out :
?mode:int -> ?flags:open_flag list ->
string -> (out_channel -> 'a) -> 'a
val write_lines_l : out_channel -> string list -> unit
----
So we see the pattern for `with_out` (which opens a function in write
mode and gives its functional argument the corresponding file descriptor).
NOTE: you should never let the resource escape the
scope of the `with_resource` call, because it will not be valid outside.
OCaml's type system doesn't make it easy to forbid that so we rely
on convention here (it would be possible, but cumbersome, using
a record with an explicitely quantified function type).
Now we can read the file again:
[source,OCaml]
----
# let lines = CCIO.with_in "/tmp/foobar" CCIO.read_lines_l ;;
val lines : string list = ["hello"; "world"]
----
There are some other functions in `CCIO` that return _generators_
instead of lists. The type of generators in containers
is `type 'a gen = unit -> 'a option` (combinators can be
found in the opam library called "gen"). A generator is to be called
to obtain successive values, until it returns `None` (which means it
has been exhausted). In particular, python users might recognize
the function
[source,OCaml]
----
# CCIO.File.walk ;;
- : string -> walk_item gen = <fun>;;
----
where `type walk_item = [ `Dir | `File ] * string` is a path
paired with a flag distinguishing files from directories.
=== To go further: containers.data
There is also a sub-library called `containers.data`, with lots of
more specialized data-structures.
The documentation contains the API for all the modules
(see link:README.adoc[the readme]); they also provide
interface to `sequence` and, as the rest of containers, minimize
dependencies over other modules. To use `containers.data` you need to link it,
either in your build system or by `#require containers.data;;`
A quick example based on purely functional double-ended queues:
[source,OCaml]
----
# #require "containers.data";;
# #install_printer CCFQueue.print;; (* better printing of queues! *)
# let q = CCFQueue.of_list [2;3;4] ;;
val q : int CCFQueue.t = queue {2; 3; 4}
# let q2 = q |> CCFQueue.cons 1 |> CCFQueue.cons 0 ;;
val q2 : int CCFQueue.t = queue {0; 1; 2; 3; 4}
(* remove first element *)
# CCFQueue.take_front q2;;
- : (int * int CCFQueue.t) option = Some (0, queue {1; 2; 3; 4})
(* q was not changed *)
# CCFQueue.take_front q;;
- : (int * int CCFQueue.t) option = Some (2, queue {3; 4})
(* take works on both ends of the queue *)
# CCFQueue.take_back_l 2 q2;;
- : int CCFQueue.t * int list = (queue {0; 1; 2}, [3; 4])
----
=== Common Type Definitions
Some structural types are used throughout the library:
gen:: `'a gen = unit -> 'a option` is an iterator type. Many combinators
are defined in the opam library https://github.com/c-cube/gen[gen]
sequence:: `'a sequence = (unit -> 'a) -> unit` is also an iterator type.
It is easier to define on data structures than `gen`, but it a bit less
powerful. The opam library https://github.com/c-cube/sequence[sequence]
can be used to consume and produce values of this type.
error:: `'a or_error = ('a, string) result = Error of string | Ok of 'a`
using the standard `result` type, supported in `CCResult`.
klist:: `'a klist = unit -> [`Nil | `Cons of 'a * 'a klist]` is a lazy list
without memoization, used as a persistent iterator. The reference
module is `CCKList` (in `containers.iter`).
printer:: `'a printer = Format.formatter -> 'a -> unit` is a pretty-printer
to be used with the standard module `Format`. In particular, in many cases,
`"foo: %a" Foo.print foo` will type-check.

View file

@ -1,275 +0,0 @@
= Tutorial
:source-highlighter: pygments
This tutorial contains a few examples to illustrate the features and
usage of containers. We assume containers is installed and that
the library is loaded, e.g. with:
[source,OCaml]
----
#require "containers";;
----
== Basics
We will start with a few list helpers, then look at other parts of
the library, including printers, maps, etc.
[source,OCaml]
----
(* quick reminder of this awesome standard operator *)
# (|>) ;;
- : 'a -> ('a -> 'b) -> 'b = <fun>
# open CCList.Infix;;
# let l = 1 -- 100;;
val l : int list = [1; 2; .....]
# l
|> CCList.filter_map
(fun x-> if x mod 3=0 then Some (float x) else None)
|> CCList.take 5 ;;
- : float list = [3.; 6.; 9.; 12.; 15.]
# let l2 = l |> CCList.take_while (fun x -> x<10) ;;
val l2 : int list = [1; 2; 3; 4; 5; 6; 7; 8; 9]
(* an extension of Map.Make, compatible with Map.Make(CCInt) *)
# module IntMap = CCMap.Make(CCInt);;
(* conversions using the "sequence" type, fast iterators that are
pervasively used in containers. Combinators can be found
in the opam library "sequence". *)
# let map =
l2
|> List.map (fun x -> x, string_of_int x)
|> CCList.to_seq
|> IntMap.of_seq;;
val map : string CCIntMap.t = <abstr>
(* check the type *)
# CCList.to_seq ;;
- : 'a list -> 'a sequence = <fun>
# IntMap.of_seq ;;
- : (int * 'a) CCMap.sequence -> 'a IntMap.t = <fun>
(* we can print, too *)
# Format.printf "@[<2>map =@ @[<hov>%a@]@]@."
(IntMap.print CCFormat.int CCFormat.string_quoted)
map;;
map =
[1 --> "1", 2 --> "2", 3 --> "3", 4 --> "4", 5 --> "5", 6 --> "6",
7 --> "7", 8 --> "8", 9 --> "9"]
- : unit = ()
(* options are good *)
# IntMap.get 3 map |> CCOpt.map (fun s->s ^ s);;
- : string option = Some "33"
----
== New types: `CCVector`, `CCHeap`, `CCError`, `CCResult`
Containers also contains (!) a few datatypes that are not from the standard
library but that are useful in a lot of situations:
CCVector::
A resizable array, with a mutability parameter. A value of type
`('a, CCVector.ro) CCVector.t` is an immutable vector of values of type `'a`,
whereas a `('a, CCVector.rw) CCVector.t` is a mutable vector that
can be modified. This way, vectors can be used in a quite functional
way, using operations such as `map` or `flat_map`, or in a more
imperative way.
CCHeap::
A priority queue (currently, leftist heaps) functorized over
a module `sig val t val leq : t -> t -> bool` that provides a type `t`
and a partial order `leq` on `t`.
CCError::
An error type for making error handling more explicit (an error monad,
really, if you're not afraid of the "M"-word). It is similar to the
more recent `CCResult`, but works with polymorphic variants for
compatibility with the numerous libraries that use the same type,
that is, `type ('a, 'b) CCError.t = [`Ok of 'a | `Error of 'b]`.
CCResult::
It uses the new `result` type from the standard library (or from
the retrocompatibility package on opam), and presents an interface
similar to `CCError`. In an indeterminate amount of time, it will
totally replace `CCError`.
Now for a few examples:
[source,OCaml]
----
(* create a new empty vector. It is mutable, for otherwise it would
not be very useful. *)
# CCVector.create;;
- : unit -> ('a, CCVector.rw) CCVector.t = <fun>
(* init, similar to Array.init, can be used to produce a
vector that is mutable OR immutable (see the 'mut parameter?) *)
# CCVector.init ;;
- : int -> (int -> 'a) -> ('a, 'mut) CCVector.t = <fun>c
(* use the infix (--) operator for creating a range. Notice
that v is a vector of integer but its mutability is not
decided yet. *)
# let v = CCVector.(1 -- 10);;
val v : (int, '_a) CCVector.t = <abstr>
# Format.printf "v = @[%a@]@." (CCVector.print CCInt.print) v;;
v = [1, 2, 3, 4, 5, 6, 7, 8, 9, 10]
(* now let's mutate v *)
# CCVector.push v 42;;
- : unit = ()
(* now v is a mutable vector *)
# v;;
- : (int, CCVector.rw) CCVector.t = <abstr>
(* functional combinators! *)
# let v2 = v
|> CCVector.map (fun x-> x+1)
|> CCVector.filter (fun x-> x mod 2=0)
|> CCVector.rev ;;
val v2 : (int, '_a) CCVector.t = <abstr>
# Format.printf "v2 = @[%a@]@." (CCVector.print CCInt.print) v2;;
v2 = [10, 8, 6, 4, 2]
(* let's transfer to a heap *)
# module IntHeap = CCHeap.Make(struct type t = int let leq = (<=) end);;
# let h = v2 |> CCVector.to_seq |> IntHeap.of_seq ;;
val h : IntHeap.t = <abstr>
(* We can print the content of h
(printing is not necessarily in order, though) *)
# Format.printf "h = [@[%a@]]@." (IntHeap.print CCInt.print) h;;
h = [2,4,6,8,10]
(* we can remove the first element, which also returns a new heap
that does not contain it — CCHeap is a functional data structure *)
# IntHeap.take h;;
- : (IntHeap.t * int) option = Some (<abstr>, 2)
# let h', x = IntHeap.take_exn h ;;
val h' : IntHeap.t = <abstr>
val x : int = 2
(* see, 2 is removed *)
# IntHeap.to_list h' ;;
- : int list = [4; 6; 8; 10]
----
== IO helpers
The core library contains a module called `CCIO` that provides useful
functions for reading and writing files. It provides functions that
make resource handling easy, following
the pattern `with_resource : resource -> (access -> 'a) -> 'a` where
the type `access` is a temporary handle to the resource (e.g.,
imagine `resource` is a file name and `access` a file descriptor).
Calling `with_resource r f` will access `r`, give the result to `f`,
compute the result of `f` and, whether `f` succeeds or raises an
error, it will free the resource.
Consider for instance:
[source,OCaml]
----
# CCIO.with_out "/tmp/foobar"
(fun out_channel ->
CCIO.write_lines_l out_channel ["hello"; "world"]);;
- : unit = ()
----
This just opened the file '/tmp/foobar', creating it if it didn't exist,
and wrote two lines in it. We did not have to close the file descriptor
because `with_out` took care of it. By the way, the type signatures are:
[source,OCaml]
----
val with_out :
?mode:int -> ?flags:open_flag list ->
string -> (out_channel -> 'a) -> 'a
val write_lines_l : out_channel -> string list -> unit
----
So we see the pattern for `with_out` (which opens a function in write
mode and gives its functional argument the corresponding file descriptor).
NOTE: you should never let the resource escape the
scope of the `with_resource` call, because it will not be valid outside.
OCaml's type system doesn't make it easy to forbid that so we rely
on convention here (it would be possible, but cumbersome, using
a record with an explicitely quantified function type).
Now we can read the file again:
[source,OCaml]
----
# let lines = CCIO.with_in "/tmp/foobar" CCIO.read_lines_l ;;
val lines : string list = ["hello"; "world"]
----
There are some other functions in `CCIO` that return _generators_
instead of lists. The type of generators in containers
is `type 'a gen = unit -> 'a option` (combinators can be
found in the opam library called "gen"). A generator is to be called
to obtain successive values, until it returns `None` (which means it
has been exhausted). In particular, python users might recognize
the function
[source,OCaml]
----
# CCIO.File.walk ;;
- : string -> walk_item gen = <fun>;;
----
where `type walk_item = [ `Dir | `File ] * string` is a path
paired with a flag distinguishing files from directories.
== To go further: containers.data
There is also a sub-library called `containers.data`, with lots of
more specialized data-structures.
The documentation contains the API for all the modules
(see link:README.adoc[the readme]); they also provide
interface to `sequence` and, as the rest of containers, minimize
dependencies over other modules. To use `containers.data` you need to link it,
either in your build system or by `#require containers.data;;`
A quick example based on purely functional double-ended queues:
[source,OCaml]
----
# #require "containers.data";;
# #install_printer CCFQueue.print;; (* better printing of queues! *)
# let q = CCFQueue.of_list [2;3;4] ;;
val q : int CCFQueue.t = queue {2; 3; 4}
# let q2 = q |> CCFQueue.cons 1 |> CCFQueue.cons 0 ;;
val q2 : int CCFQueue.t = queue {0; 1; 2; 3; 4}
(* remove first element *)
# CCFQueue.take_front q2;;
- : (int * int CCFQueue.t) option = Some (0, queue {1; 2; 3; 4})
(* q was not changed *)
# CCFQueue.take_front q;;
- : (int * int CCFQueue.t) option = Some (2, queue {3; 4})
(* take works on both ends of the queue *)
# CCFQueue.take_back_l 2 q2;;
- : int CCFQueue.t * int list = (queue {0; 1; 2}, [3; 4])
----