Low level IO abstraction. A typical implementation is unix. This abstraction is meant to be dead simple. Not a lot of documentation is required.
It is not resistant to race condictions. There should not be concurrent modifications of the files.
These functions are essentially invoking the underlying functions from Unix directly; there is no buffering for example.
Errors
type misc_error = Unix.error * string * stringAn abstract error type that contains the IO-backend specific errors. (e.g. Unix.error)
type create_error = [ | `Io_misc of misc_error| `File_exists of string
]type open_error = [ | `Io_misc of misc_error| `No_such_file_or_directory| `Not_a_file
]type read_error = [ | `Io_misc of misc_error| `Read_out_of_bounds| `Closed| `Invalid_argument
]type write_error = [ | `Io_misc of misc_error| `Ro_not_allowed| `Closed
]type close_error = [ | `Io_misc of misc_error| `Double_close
]type mkdir_error = [ | `Io_misc of misc_error| `File_exists of string| `No_such_file_or_directory| `Invalid_parent_directory
]Safe Functions
None of the functions in this section raise exceptions. They may however perform effects that are always continued.
Life Cycle
val create :
path:string ->
overwrite:bool ->
(t, [> create_error ]) Stdlib.resultval open_ : path:string -> readonly:bool -> (t, [> open_error ]) Stdlib.resultWrite Functions
val write_string :
t ->
off:Optint.Int63.t ->
string ->
(unit, [> write_error ]) Stdlib.resultwrite_string t ~off s writes s at offset in t.
fsync t persists to the file system the effects of previous create or write.
val move_file :
src:string ->
dst:string ->
(unit, [> `Sys_error of string ]) Stdlib.resultval copy_file :
src:string ->
dst:string ->
(unit, [> `Sys_error of string ]) Stdlib.resultval mkdir : string -> (unit, [> mkdir_error ]) Stdlib.resultval unlink : string -> (unit, [> `Sys_error of string ]) Stdlib.resultRead Functions
val read_to_string :
t ->
off:Optint.Int63.t ->
len:int ->
(string, [> read_error ]) Stdlib.resultread_to_string t ~off ~len are the len bytes of t at off.
val read_all_to_string :
t ->
(string, [> `Io_misc of misc_error | `Closed ]) Stdlib.resultread_to_string t is the contents full contents of the file.
The individual pages are not guaranteed to be read atomically.
val read_size : t -> (Optint.Int63.t, [> read_error ]) Stdlib.resultread_size t is the number of bytes of the file handled by t.
This function is expensive in the unix implementation because it performs syscalls.
val size_of_path :
string ->
(Optint.Int63.t,
[> `Io_misc of misc_error | `No_such_file_or_directory | `Not_a_file ])
Stdlib.resultval classify_path :
string ->
[> `File | `Directory | `No_such_file_or_directory | `Other ]MISC.
Unsafe Functions
These functions are equivalents to exising safe ones, but using exceptions instead of the result monad for performances reasons.
val read_exn : t -> off:Optint.Int63.t -> len:int -> bytes -> unitread_exn t ~off ~len b reads the len bytes of t at off to b.
Raises Errors.Pack_error and Errors.RO_not_allowed.
Also raises backend-specific exceptions (e.g. Unix.Unix_error for the unix backend).
val write_exn : t -> off:Optint.Int63.t -> len:int -> string -> unitwrite_exn t ~off ~len b writes the first len bytes pf b to t at offset off.
Raises Errors.Pack_error and Errors.RO_not_allowed.
Also raises backend-specific exceptions (e.g. Unix.Unix_error for the unix backend).
val catch_misc_error :
(unit -> 'a) ->
('a, [> `Io_misc of misc_error ]) Stdlib.result