A modular, clean and powerful extension of the OCaml standard library.
(Jump to the current API documentation)
Containers is an extension of OCaml's standard library (under BSD license)
focused on data structures, combinators and iterators, without dependencies on
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.
Containers is:
- A usable, reasonably well-designed library that extends OCaml's standard
library (in 'src/core/', packaged under
containers
in ocamlfind. Modules are totally independent and are prefixed withCC
(for "containers-core" or "companion-cube" because I'm a megalomaniac). This part should be usable and should work. For instance,CCList
contains functions and lists including safe versions ofmap
andappend
. It also provides a drop-in replacement to the standard library, in the moduleContainers
(intended to be opened, replaces some stdlib modules with extended ones), and a small S-expression printer and parser that can be functorized over the representation of values. - Some sub-libraries with a specific focus each:
- Utilities around the
unix
library incontainers.unix
(mainly to spawn sub-processes easily and deal with resources safely) - A bencode codec in
containers.bencode
. This is a tiny json-like serialization format that is extremely simple. It comes from bittorrent files. - A CBOR codec in
containers.cbor
. This is a compact binary serialization format. - The Strongly Connected Component
algorithm, functorized, in
containers.scc
- Utilities around the
- A separate library
containers-data
with additional data structures that don't have an equivalent in the standard library, typically not as thoroughly maintained. This is now in its own package since 3.0.
Some of the modules have been moved to their own repository (e.g. sequence
(now iter
),
gen
, qcheck
) and are on opam for great fun and profit.
Containers-thread has been removed in favor of Moonpool.
The changelog's breaking section contains a list of the breaking changes in this release.
- The biggest change is that some sub-libraries have been either turned into
their own packages (
containers-data
), deleted (containers.iter
),or merged elsewhere (containers.sexp
). This means that if use these libraries you will have to edit yourdune
/_oasis
/opam
files.
- if you use
containers.sexp
(i.e. theCCSexp
module), it now lives incontainers
itself. - if you used anything in
containers.data
, you need to depend on thecontainers-data
package now.
- Another large change is the removal (at last!) of functions deprecated
in 2.8, related to the spread of
Seq.t
as the standard iterator type. Functions likeCCVector.of_seq
now operate on this standardSeq.t
type, and old-time iteration based on iter is now namedof_iter
,to_iter
, etc.
Here you need to change your code, possibly using search and replace. Thankfully, the typechecker should guide you.
-
Array_slice
andString.Sub
have been removed to simplify the code andString
more lightweight. There is no replacement at the moment. Please tell us if you need this to be turned into a sub-library. -
Renaming of some functions into more explicit/clear names. Examples:
CCVector.shrink
is nowCCVector.truncate
CCVector.remove
is nowCCVector.remove_unordered
, to be contrasted with the newCCVector.remove_and_shift
.CCPair.map_fst
andmap_snd
now transform a tuple into another tuple by modify the first (resp. second) element.
-
All the collection pretty-printers now take their separator/start/stop optional arguments as
unit printer
(i.e.Format.formatter -> unit -> unit
functions) rather than strings. This gives the caller better control over the formatting of lists, arrays, queues, tables, etc. -
Removal of many deprecated functions.
-
The type system should detect issues related to
print
renamed intopp
easily. If you are lucky, a call tosed -i 's/print/pp/g'
on the concerned files might help rename all the calls properly. -
many optional arguments have become mandatory, because their default value would be a polymorphic "magic" operator such as
(=)
or(>=)
. Now these have to be specified explicitly, but during the transition you can useStdlib.(=)
andStdlib.(>=)
as explicit arguments. -
if your code contains
open Containers
, the biggest hurdle you face might be that operators have become monomorphic by default. We believe this is a useful change that prevents many subtle bugs. However, during migration and until you use proper combinators for equality (CCEqual
), comparison (CCOrd
), and hashing (CCHash
), you might want to addopen Stdlib
just after theopen Containers
. See the section on monomorphic operators for more details.
To quote @bluddy in #196:
The main problem with polymorphic comparison is that many data structures will give one result for structural comparison, and a different result for semantic comparison. The classic example is comparing maps. If you have a list of maps and try to use comparison to sort them, you'll get the wrong result: multiple map structures can represent the same semantic mapping from key to value, and comparing them in terms of structure is simply wrong. A far more pernicious bug occurs with hashtables. Identical hashtables will seem to be identical for a while, as before they've had a key clash, the outer array is likely to be the same. Once you get a key clash though, you start getting lists inside the arrays (or maps inside the arrays if you try to make a smarter hashtable) and that will cause comparison errors ie. identical hashtables will be seen as different or vice versa.
Every time you use a polymorphic comparison where you're using a data type where structural comparison != semantic comparison, it's a bug. And every time you use polymorphic comparison where the type of data being compared may vary (e.g. it's an int now, but it may be a map later), you're planting a bug for the future.
See also:
- https://blog.janestreet.com/the-perils-of-polymorphic-compare/
- https://blog.janestreet.com/building-a-better-compare/
If you just want to use polymorphic operators, it's fine! You can access them
easily by using Stdlib.(=)
, Stdlib.max
, etc.
When migrating a module, you can add open Stdlib
on top of it to restore
the default behavior. It is, however, recommended to export an equal
function
(and compare
, and hash
) for all the public types, even if their internal
definition is just the corresponding polymorphic operator.
This way, other modules can refer to Foo.equal
and will not have to be
updated the day Foo.equal
is no longer just polymorphic equality.
Another bonus is that Hashtbl.Make(Foo)
or Map.Make(Foo)
will just work™.
To print values with types defined in containers
in the bytecode debugger,
you first have to load the appropriate bytecode archives. After starting a
session, e.g. ocamldebug your_program.bc
,
# #load_printer containers_monomorphic.cma;;
# #load_printer containers.cma;;
For these archives to be found, you may have to run
the program first. Now
printing functions that have the appropriate type Format.formatter -> 'a -> unit
can be installed. For example,
# #install_printer Containers.Int.pp;;
However, printer combinators are not easily handled by ocamldebug
. For
instance # install_printer Containers.(List.pp Int.pp)
will not work out of
the box. You can make this work by writing a short module which defines
ready-made combined printing functions, and loading that in ocamldebug. For
instance
module M = struct
let pp_int_list = Containers.(List.pp Int.pp)
end;;
loaded via # load_printer m.cmo
and installed as # install_printer M.pp_int_list
.
See this file.
- Mailing List the address is mailto:[email protected]
- the github wiki
- on IRC, ask
companion_cube
on#[email protected]
- there is a
#containers
channel on OCaml's discord server.
You might start with the tutorial to get a picture of how to use the library.
You can either build and install the library (see build), or just copy files to your own project. The last solution has the benefits that you don't have additional dependencies nor build complications (and it may enable more inlining). Since modules have a friendly license and are mostly independent, both options are easy.
In a toplevel, using ocamlfind:
# #use "topfind";;
...
# #require "containers";;
# #require "containers-data";;
# CCList.flat_map;;
- : ('a -> 'b list) -> 'a list -> 'b list = <fun>
# open Containers (* optional *);;
# List.flat_map ;;
- : ('a -> 'b list) -> 'a list -> 'b list = <fun>
If you have comments, requests, or bugfixes, please share them! :-)
This code is free, under the BSD license.
See the documentation and the tutorial below for a gentle introduction.
In general, see http://c-cube.github.io/ocaml-containers/last/ for the API documentation.
Some examples can be found there, per-version doc there.
You will need OCaml >=
4.03.0.
The preferred way to install is through opam.
$ opam install containers
You need dune (formerly jbuilder).
$ make
To build and run tests (requires qcheck-core
, gen
, iter
):
$ opam install qcheck-core
$ make test
To build the small benchmarking suite (requires benchmark):
$ opam install benchmark batteries
$ make bench
$ ./benchs/run_benchs.sh
PRs on github are very welcome (patches by email too, if you prefer so).
how to contribute (click to unfold)
The list of contributors can be seen on github.
Alternatively, git authors
from git-extras can be invoked from within the repo
to list authors based on the git commits.
Assuming your are in a clone of the repository:
- Some dependencies are required, you'll need
opam install benchmark qcheck-core iter gen mdx uutf yojson
. - run
make all
to enable everything (including tests). - make your changes, commit, push, and open a PR.
- use
make test
without moderation! It must pass before a PR is merged. There are around 1150 tests right now, and new features should come with their own tests.
If you feel like writing new tests, that is totally worth a PR (and my gratefulness).
A few guidelines to follow the philosophy of containers:
- no dependencies between basic modules (even just for signatures);
- add
@since
tags for new functions; - add tests if possible (see
tests/
dir) There are numerous inline tests already, to see what it looks like search for comments starting with(*$
in source files.
Thanks for wanting to contribute! To contribute a change, here are the steps (roughly):
-
click "fork" on https://github.com/c-cube/ocaml-containers on the top right of the page. This will create a copy of the repository on your own github account.
-
click the big green "clone or download" button, with "SSH". Copy the URL (which should look like
[email protected]:<your username>/ocaml-containers.git
) into a terminal to enter the command:$ git clone [email protected]:<your username>/ocaml-containers.git
-
then,
cd
into the newly created directory. -
make the changes you want. See <#first-time-contributors> for more details about what to do in particular.
-
use
git add
andgit commit
to commit these changes. -
git push origin master
to push the new change(s) onto your copy of the repository -
on github, open a "pull request" (PR). Et voilà !
This tutorial contains a few examples to illustrate the features and usage of containers.
an introduction to containers (click to unfold)
We assume containers is installed and that the library is loaded, e.g. with:
# #require "containers";;
# Format.set_margin 50 (* for readability here *);;
- : unit = ()
We will start with a few list helpers, then look at other parts of the library, including printers, maps, etc.
# (|>) (* quick reminder of this awesome standard operator *);;
- : 'a -> ('a -> 'b) -> 'b = <fun>
# 10 |> succ;;
- : int = 11
# open CCList.Infix;;
# let l = 1 -- 100;;
val l : int list =
[1; 2; 3; 4; 5; 6; 7; 8; 9; 10; 11; 12; 13; 14; 15; 16; 17; 18; 19; 20; 21;
22; 23; 24; 25; 26; 27; 28; 29; 30; 31; 32; 33; 34; 35; 36; 37; 38; 39;
40; 41; 42; 43; 44; 45; 46; 47; 48; 49; 50; 51; 52; 53; 54; 55; 56; 57;
58; 59; 60; 61; 62; 63; 64; 65; 66; 67; 68; 69; 70; 71; 72; 73; 74; 75;
76; 77; 78; 79; 80; 81; 82; 83; 84; 85; 86; 87; 88; 89; 90; 91; 92; 93;
94; 95; 96; 97; 98; 99; 100]
# (* transform a list, dropping some elements *)
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 "iter" type, fast iterators that are
pervasively used in containers. Combinators can be found
in the opam library "iter". *)
let map : string IntMap.t =
l2
|> List.map (fun x -> x, string_of_int x)
|> CCList.to_iter
|> IntMap.of_iter;;
val map : string IntMap.t = <abstr>
# CCList.to_iter (* check the type *);;
- : 'a list -> 'a CCList.iter = <fun>
# IntMap.of_iter ;;
- : (int * 'a) CCMap.iter -> 'a IntMap.t = <fun>
# (* we can print, too *)
Format.printf "@[<2>map =@ @[<hov>%a@]@]@."
(IntMap.pp 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 |> CCOption.map (fun s->s ^ s);;
- : string option = Some "33"
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 asmap
orflat_map
, or in a more imperative way.CCHeap
: A priority queue (currently, leftist heaps) functorized over a modulesig val t val leq : t -> t -> bool
that provides a typet
and a partial orderleq
ont
.CCResult
An error type for making error handling more explicit (an error monad, really, if you're not afraid of the "M"-word). Subsumes and replaces the oldCCError
. It uses the newresult
type from the standard library (or from the retrocompatibility package on opam) and provides many combinators for dealing withresult
.CCSexp
andCCCanonical_sexp
: functorized printer and parser for S-expressions, respectively as actual S-expressions (likesexplib
) and as canonical binary-safe S-expressions (likecsexp
)CCByte_buffer
: a better version of the standardBuffer.t
which cannot be extended and prevents access to its internal byte array. This type is designed for (blocking) IOs and to produce complex strings incrementally in an efficient way.
Now for a few examples:
# (* 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>
# (* 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.pp CCInt.pp) v;;
v = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
- : unit = ()
# CCVector.push v 42;;
- : unit = ()
# v (* now v is a mutable vector *);;
- : (int, CCVector.rw) CCVector.t = <abstr>
# (* functional combinators! *)
let v2 : _ CCVector.ro_vector = v
|> CCVector.map (fun x-> x+1)
|> CCVector.filter (fun x-> x mod 2=0)
|> CCVector.rev ;;
val v2 : int CCVector.ro_vector = <abstr>
# Format.printf "v2 = @[%a@]@." (CCVector.pp CCInt.pp) v2;;
v2 = 10, 8, 6, 4, 2
- : unit = ()
(* let's transfer to a heap *)
module IntHeap = CCHeap.Make(struct type t = int let leq = (<=) end);;
# let h = v2 |> CCVector.to_iter |> IntHeap.of_iter ;;
val h : IntHeap.t = <abstr>
# (* We can print the content of h
(printing is not necessarily in order, though) *)
Format.printf "h = [@[%a@]]@." (IntHeap.pp CCInt.pp) h;;
h = [2,4,6,8,10]
- : unit = ()
# (* 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
# IntHeap.to_list h' (* see, 2 is removed *);;
- : int list = [4; 6; 8; 10]
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:
# CCIO.with_out "./foobar"
(fun out_channel ->
CCIO.write_lines_l out_channel ["hello"; "world"]);;
- : unit = ()
This just opened the file '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:
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 explicitly quantified function type).
Now we can read the file again:
# let lines : string list = CCIO.with_in "./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
# 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.
There is also a library called containers-data
, with lots of
more specialized data-structures.
The documentation contains the API for all the modules; they also provide
interface to iter
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:
# #require "containers-data";;
# #install_printer CCFQueue.pp (* 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])
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 gen -
iter
:'a iter = (unit -> 'a) -> unit
is also an iterator type, formerly namedsequence
. It is easier to define on data structures thangen
, but it a bit less powerful. The opam library iter can be used to consume and produce values of this type.It was renamed from
'a sequence
to'a iter
to distinguish it better fromCore.Sequence
and the standardseq
. -
error
:'a or_error = ('a, string) result = Error of string | Ok of 'a
using the standardresult
type, supported inCCResult
. -
printer
:'a printer = Format.formatter -> 'a -> unit
is a pretty-printer to be used with the standard moduleFormat
. In particular, in many cases,"foo: %a" Foo.print foo
will type-check.
See the extended documentation for more examples.
Beforehand, check grep deprecated -r src
to see whether some functions
can be removed.
make all
- update version in
containers.opam
make update_next_tag
(to update@since
comments; be careful not to change symlinks)- check status of modules (
{b status: foo}
) and update if required; removed deprecated functions, etc. - update
CHANGELOG.md
(see its end to find the right git command) - commit the changes
make test doc
export VERSION=<tag here>; git tag -f $VERSION; git push origin :$VERSION; git push origin $VERSION
- new opam package:
opam publish https://github.com/c-cube/ocaml-containers/archive/<tag>.tar.gz
- re-generate doc:
make doc
and put it intogh-pages
git log --format='%aN' | sort -u