file

file

File Interface Module

The module file provides an interface to the file system.

On operating systems with thread support (Solaris and Windows), it is possible to let file operations be performed in threads of their own, allowing other Erlang processes to continue executing in parallel with the file operations. See the command line flag +A in erl(1).

iodata() = iolist() | binary()
  iolist() = [char() | binary() | iolist()]

io_device()
  as returned by file:open/2, a process handling IO protocols

name() = string() | atom() | DeepList
  DeepList = [char() | atom() | DeepList]

posix()
  an atom which is named from the Posix error codes used in
  Unix, and in the runtime libraries of most C compilers

time() = {{Year, Month, Day}, {Hour, Minute, Second}}
  Year = Month = Day = Hour = Minute = Second = int()
  Must denote a valid date and time
    

Types:

Changes group of a file. See write_file_info/2.

Types:

Changes owner of a file. See write_file_info/2.

Types:

Changes owner and group of a file. See write_file_info/2.

Types:

Changes the modification and access times of a file. See write_file_info/2.

Types:

Changes the modification and last access times of a file. See write_file_info/2.

Types:

Closes the file referenced by IoDevice. It mostly returns ok, expect for some severe errors such as out of memory.

Note that if the option delayed_write was used when opening the file, close/1 might return an old write error and not even try to close the file. See open/2.

Types:

Reads Erlang terms, separated by '.', from Filename. Returns one of the following:

{ok, Terms}

The file was successfully read.

{error, Posix}

An error occurred when opening the file or reading it. See open/2 for a list of typical error codes.

{error, {Line, Mod, Term}}

An error occurred when interpreting the Erlang terms in the file. Use format_error/1 to convert the three-element tuple to an English description of the error.

Example:

f.txt:  {person, "kalle", 25}.
        {person, "pelle", 30}.
        

1> file:consult("f.txt").
{ok,[{person,"kalle",25},{person,"pelle",30}]}
        

Types:

Copies ByteCount bytes from Source to Destination. Source and Destination refer to either filenames or IO devices from e.g. open/2. ByteCount defaults infinity, denoting an infinite number of bytes.

The argument Modes is a list of possible modes, see open/2, and defaults to [].

If both Source and Destination refer to filenames, the files are opened with [read, binary] and [write, binary] prepended to their mode lists, respectively, to optimize the copy.

If Source refers to a filename, it is opened with read mode prepended to the mode list before the copy, and closed when done.

If Destination refers to a filename, it is opened with write mode prepended to the mode list before the copy, and closed when done.

Returns {ok, BytesCopied} where BytesCopied is the number of bytes that actually was copied, which may be less than ByteCount if end of file was encountered on the source. If the operation fails, {error, Reason} is returned.

Typical error reasons: As for open/2 if a file had to be opened, and as for read/2 and write/2.

Types:

Tries to delete the directory Dir. The directory must be empty before it can be deleted. Returns ok if successful.

Typical error reasons are:

eacces

Missing search or write permissions for the parent directories of Dir.

eexist

The directory is not empty.

enoent

The directory does not exist.

enotdir

A component of Dir is not a directory. On some platforms, enoent is returned instead.

einval

Attempt to delete the current directory. On some platforms, eacces is returned instead.

Types:

Tries to delete the file Filename. Returns ok if successful.

Typical error reasons are:

enoent

The file does not exist.

eacces

Missing permission for the file or one of its parents.

eperm

The file is a directory and the user is not super-user.

enotdir

A component of the file name is not a directory. On some platforms, enoent is returned instead.

Types:

Reads and evaluates Erlang expressions, separated by '.' (or ',', a sequence of expressions is also an expression), from Filename. The actual result of the evaluation is not returned; any expression sequence in the file must be there for its side effect. Returns one of the following:

ok

The file was read and evaluated.

{error, Posix}

An error occurred when opening the file or reading it. See open/2 for a list of typical error codes.

{error, {Line, Mod, Term}}

An error occurred when interpreting the Erlang expressions in the file. Use format_error/1 to convert the three-element tuple to an English description of the error.

Types:

The same as eval/1 but the variable bindings Bindings are used in the evaluation. See erl_eval(3) about variable bindings.

This function is obsolete. Use read_file_info/1 instead.

Types:

Given the error reason returned by any function in this module, returns a descriptive string of the error in English.

Types:

Returns {ok, Dir}, where Dir is the current working directory of the file server.

In rare circumstances, this function can fail on Unix. It may happen if read permission does not exist for the parent directories of the current directory.

Typical error reasons are:

eacces

Missing read permission for one of the parents of the current directory.

Types:

Drive should be of the form "Letter:", for example "c:". Returns {ok, Dir} or {error, Reason}, where Dir is the current working directory of the drive specified.

This function returns {error, enotsup} on platforms which have no concept of current drive (Unix, for example).

Typical error reasons are:

enotsup

The operating system have no concept of drives.

eacces

The drive does not exist.

einval

The format of Drive is invalid.

Types:

Lists all the files in a directory. Returns {ok, Filenames} if successful. Otherwise, it returns {error, Reason}. Filenames is a list of the names of all the files in the directory. The names are not sorted.

Typical error reasons are:

eacces

Missing search or write permissions for Dir or one of its parent directories.

enoent

The directory does not exist.

Types:

Tries to create the directory Dir. Missing parent directories are not created. Returns ok if successful.

Typical error reasons are:

eacces

Missing search or write permissions for the parent directories of Dir.

eexist

There is already a file or directory named Dir.

enoent

A component of Dir does not exist.

enospc

There is a no space left on the device.

enotdir

A component of Dir is not a directory. On some platforms, enoent is returned instead.

Types:

Makes a hard link from Existing to New, on platforms that support links (Unix). This function returns ok if the link was successfully created, or {error, Reason}. On platforms that do not support links, {error,enotsup} is returned.

Typical error reasons:

eacces

Missing read or write permissions for the parent directories of Existing or New.

eexist

New already exists.

enotsup

Hard links are not supported on this platform.

Types:

This function creates a symbolic link Name2 to the file or directory Name1, on platforms that support symbolic links (most Unix systems). Name1 need not exist. This function returns ok if the link was successfully created, or {error, Reason}. On platforms that do not support symbolic links, {error, enotsup} is returned.

Typical error reasons:

eacces

Missing read or write permissions for the parent directories of Name1 or Name2.

eexist

Name2 already exists.

enotsup

Symbolic links are not supported on this platform.

Types:

Opens the file Filename in the mode determined by Modes, which may contain one or more of the following items:

read

The file, which must exist, is opened for reading.

write

The file is opened for writing. It is created if it does not exist. If the file exists, and if write is not combined with read, the file will be truncated.

append

The file will be opened for writing, and it will be created if it does not exist. Every write operation to a file opened with append will take place at the end of the file.

raw

The raw option allows faster access to a file, because no Erlang process is needed to handle the file. However, a file opened in this way has the following limitations:

• The functions in the io module cannot be used, because they can only talk to an Erlang process. Instead, use the read/2 and write/2 functions.
• Only the Erlang process which opened the file can use it.
• A remote Erlang file server cannot be used; the computer on which the Erlang node is running must have access to the file system (directly or through NFS).

binary

This option can only be used if the raw option is specified as well. When specified, read operations on the file using the read/2 function will return binaries rather than lists.

{delayed_write, Size, Delay}

If this option is used, the data in subsequent write/2 calls is buffered until there are at least Size bytes buffered, or until the oldest buffered data is Delay milliseconds old. Then all buffered data is written in one operating system call. The buffered data is also flushed before some other file operation than write/2 is executed.
The purpose of this option is to increase performance by reducing the number of operating system calls, so the write/2 calls should be for sizes significantly less than Size, and not interspersed by to many other file operations, for this to happen.
When this option is used, the result of write/2 calls may prematurely be reported as successful, and if a write error should actually occur the error is reported as the result of the next file operation, which is not executed.
For example, when delayed_write is used, after a number of write/2 calls, close/1 might return {error, enospc} because there was not enough space on the disc for previously written data, and close/1 should probably be called again since the file is still open.

delayed_write

The same as {delayed_write, Size, Delay} with reasonable default values for Size and Delay. (Roughly some 64 KBytes, 2 seconds)

{read_ahead, Size}

This option activates read data buffering. If read/2 calls are for significantly less than Size bytes, read operations towards the operating system are still performed for blocks of Size bytes. The extra data is buffered and returned in subsequent read/2 calls, giving a performance gain since the number of operating system calls is reduced.
If read/2 calls are for sizes not significantly less than, or even greater than Size bytes, no performance gain can be expected.

read_ahead

The same as {read_ahead, Size} with a reasonable default value for Size. (Roughly some 64 KBytes)

compressed

Makes it possible to read and write gzip compressed files. Note that the file size obtained with read_file_info/1 will most probably not match the number of bytes that can be read from a compressed file.

Returns:

{ok, IoDevice}

The file has been opened in the requested mode. IoDevice is a reference to the file.

{error, Reason}

The file could not be opened.

IoDevice is really the pid of the process which handles the file. This process is linked to the process which originally opened the file. If any process to which the IoDevice is linked terminates, the file will be closed and the process itself will be terminated. An IoDevice returned from this call can be used as an argument to the IO functions (see io(3)).

In previous versions of file, modes were given as one of the atoms read, write, or read_write instead of a list. This is still allowed for reasons of backwards compatibility, but should not be used for new code. Also note that read_write is not allowed in a mode list.

Typical error reasons:

enoent

The file does not exist.

eacces

Missing permission for reading the file or searching one of the parent directories.

eisdir

The named file is not a regular file. It may be a directory, a fifo, or a device.

enotdir

A component of the file name is not a directory. On some platforms, enoent is returned instead.

enospc

There is a no space left on the device (if write access was specified).

Types:

Searches the path Path (a list of directory names) until the file Filename is found. If Filename is an absolute filename, Path is ignored. Then reads Erlang terms, separated by '.', from the file. Returns one of the following:

{ok, Terms, FullName}

The file was successfully read. FullName is the full name of the file.

{error, enoent}

The file could not be found in any of the directories in Path.

{error, Posix}

An error occurred when opening the file or reading it. See open/2 for a list of typical error codes.

{error, {Line, Mod, Term}}

An error occurred when interpreting the Erlang terms in the file. Use format_error/1 to convert the three-element tuple to an English description of the error.

Types:

Searches the path Path (a list of directory names) until the file Filename is found. If Filename is an absolute file name, Path is ignored. Then reads and evaluates Erlang expressions, separated by '.' (or ',', a sequence of expressions is also an expression), from the file. The actual result of evaluation is not returned; any expression sequence in the file must be there for its side effect. Returns one of the following:

{ok, FullName}

The file was read and evaluated. FullName is the full name of the file.

{error, enoent}

The file could not be found in any of the directories in Path.

{error, Posix}

An error occurred when opening the file or reading it. See open/2 for a list of typical error codes.

{error, {Line, Mod, Term}}

An error occurred when interpreting the Erlang expressions in the file. Use format_error/1 to convert the three-element tuple to an English description of the error.

Types:

Searches the path Path (a list of directory names) until the file Filename is found. If Filename is an absolute file name, Path is ignored. Then opens the file in the mode determined by Modes. Returns one of the following:

{ok, IoDevice, FullName}

The file has been opened in the requested mode. IoDevice is a reference to the file and FullName is the full name of the file.

{error, enoent}

The file could not be found in any of the directories in Path.

{error, Posix}

The file could not be opened.

Types:

Searches the path Path (a list of directory names) until the file Filename is found. If Filename is an absolute file name, Path is ignored. Then reads and evalutates Erlang expressions, separated by '.' (or ',', a sequence of expressions is also an expression), from the file. Returns one of the following:

{ok, Value, FullName}

The file was read and evaluated. FullName is the full name of the file and Value the value of the last expression.

{error, enoent}

The file could not be found in any of the directories in Path.

{error, Posix}

An error occurred when opening the file or reading it. See open/2 for a list of typical error codes.

{error, {Line, Mod, Term}}

An error occurred when interpreting the Erlang expressions in the file. Use format_error/1 to convert the three-element tuple to an English description of the error.

Types:

The same as path_script/2 but the variable bindings Bindings are used in the evaluation. See erl_eval(3) about variable bindings.

Types:

If Pid is an IO device, that is, a pid returned from open/2, this function returns the filename, or rather:

{ok, Filename}

If this node's file server is not a slave, the file was opened by this node's file server, (this implies that Pid must be a local pid) and the file is not closed. Filename is the filename in flat string format.

undefined

In all other cases.

This function is intended for debugging only.

Types:

Sets the position of the file referenced by IoDevice to Location. Returns {ok, NewPosition} (as absolute offset) if successful, otherwise {error, Reason}. Location is one of the following:

Offset

The same as {bof, Offset}.

{bof, Offset}

Absolute offset.

{cur, Offset}

Offset from the current position.

{eof, Offset}

Offset from the end of file.

bof | cur | eof

The same as above with Offset 0.

Typical error reasons are:

einval

Either Location was illegal, or it evaluated to a negative offset in the file. Note that if the resulting position is a negative value, the result is an error, and after the call the file position is undefined.

Types:

Performs a sequence of pread/3 in one operation, which is more efficient than calling them one at a time. Returns {ok, [Data, ...]} or {error, Reason}, where each Data, the result of the corresponding pread, is either a list or a binary depending on the mode of the file, or eof if the requested position was beyond end of file.

Types:

Combines position/2 and read/2 in one operation, which is more efficient than calling them one at a time. If IoDevice has been opened in raw mode, some restrictions apply: Location is only allowed to be an integer; and the current position of the file is undefined after the operation.

Types:

Performs a sequence of pwrite/3 in one operation, which is more efficient than calling them one at a time. Returns ok or {error, {N, Reason}}, where N is the number of successful writes that was done before the failure.

Types:

Combines position/2 and write/2 in one operation, which is more efficient than calling them one at a time. If IoDevice has been opened in raw mode, some restrictions apply: Location is only allowed to be an integer; and the current position of the file is undefined after the operation.

Types:

Reads Number bytes from the file referenced by IoDevice. This function is the only way to read from a file opened in raw mode (although it works for normally opened files, too). Returns:

{ok, Data}

If the file was opened in binary mode, the read bytes are returned in a binary, otherwise in a list. The list or binary will be shorter than the number of bytes requested if end of file was reached.

eof

Returned if Number>0 and end of file was reached before anything at all could be read.

{error, Reason}

An error occurred.

Typical error reasons:

ebadf

The file is not opened for reading.

Types:

Returns {ok, Binary}, where Binary is a binary data object that contains the contents of Filename, or {error, Reason} if an error occurs.

Typical error reasons:

enoent

The file does not exist.

eacces

Missing permission for reading the file, or for searching one of the parent directories.

eisdir

The named file is a directory.

enotdir

A component of the file name is not a directory. On some platforms, enoent is returned instead.

enomem

There is not enough memory for the contents of the file.

Types:

Retrieves information about a file. Returns {ok, FileInfo} if successful, otherwise {error, Reason}. FileInfo is a record file_info, defined in the Kernel include file file.hrl. Include the following directive in the module from which the function is called:

-include_lib("kernel/include/file.hrl").
        

The record file_info contains the following fields.

size = int()

Size of file in bytes.

type = device | directory | regular | other

The type of the file.

access = read | write | read_write | none

The current system access to the file.

atime = time()

The last (local) time the file was read.

mtime = time()

The last (local) time the file was written.

ctime = time()

The interpretation of this time field depends on the operating system. On Unix, it is the last time the file or the inode was changed. In Windows, it is the create time.

mode = int()

The file permissions as the sum of the following bit values:

8#00400

read permission: owner

8#00200

write permission: owner

8#00100

execute permission: owner

8#00040

read permission: group

8#00020

write permission: group

8#00010

execute permission: group

8#00004

read permission: other

8#00002

write permission: other

8#00001

execute permission: other

16#800

set user id on execution

16#400

set group id on execution

On Unix platforms, other bits than those listed above may be set.

links = int()

Number of links to the file (this will always be 1 for file systems which have no concept of links).

major_device = int()

Identifies the file system where the file is located. In Windows, the number indicates a drive as follows: 0 means A:, 1 means B:, and so on.

minor_device = int()

Only valid for character devices on Unix. In all other cases, this field is zero.

inode = int()

Gives the inode number. On non-Unix file systems, this field will be zero.

uid = int()

Indicates the owner of the file. Will be zero for non-Unix file systems.

gid = int()

Gives the group that the owner of the file belongs to. Will be zero for non-Unix file systems.

Typical error reasons:

eacces

Missing search permission for one of the parent directories of the file.

enoent

The file does not exist.

enotdir

A component of the file name is not a directory. On some platforms, enoent is returned instead.

Types:

This function returns {ok, Filename} if Name refers to a symbolic link or {error, Reason} otherwise. On platforms that do not support symbolic links, the return value will be {error,enotsup}.

Typical error reasons:

einval

Linkname does not refer to a symbolic link.

enoent

The file does not exist.

enotsup

Symbolic links are not supported on this platform.

Types:

This function works like read_file_info/1, except that if Name is a symbolic link, information about the link will be returned in the file_info record and the type field of the record will be set to symlink.

If Name is not a symbolic link, this function returns exactly the same result as read_file_info/1. On platforms that do not support symbolic links, this function is always equvivalent to read_file_info/1.

Types:

Tries to rename the file Source to Destination. It can be used to move files (and directories) between directories, but it is not sufficient to specify the destination only. The destination file name must also be specified. For example, if bar is a normal file and foo and baz are directories, rename("foo/bar", "baz") returns an error, but rename("foo/bar", "baz/bar") succeeds. Returns ok if it is successful.

Renaming of open files is not allowed on most platforms (see eacces below).

Typical error reasons:

eacces

Missing read or write permissions for the parent directories of Source or Destination. On some platforms, this error is given if either Source or Destination is open.

eexist

Destination is not an empty directory. On some platforms, also given when Source and Destination are not of the same type.

einval

Source is a root directory, or Destination is a sub-directory of Source.

eisdir

Destination is a directory, but Source is not.

enoent

Source does not exist.

enotdir

Source is a directory, but Destination is not.

exdev

Source and Destination are on different file systems.

Types:

Reads and evaluates Erlang expressions, separated by '.' (or ',', a sequence of expressions is also an expression), from the file. Returns one of the following:

{ok, Value}

The file was read and evaluated. Value is the value of the last expression.

{error, Posix}

An error occurred when opening the file or reading it. See open/2 for a list of typical error codes.

{error, {Line, Mod, Term}}

An error occurred when interpreting the Erlang expressions in the file. Use format_error/1 to convert the three-element tuple to an English description of the error.

Types:

The same as script/1 but the variable bindings Bindings are used in the evaluation. See erl_eval(3) about variable bindings.

Types:

Sets the current working directory of the file server to Dir. Returns ok if successful.

Typical error reasons are:

enoent

The directory does not exist.

enotdir

A component of Dir is not a directory. On some platforms, enoent is returned.

eacces

Missing permission for the directory or one of its parents.

Types:

Makes sure that any buffers kept by the operating system (not by the Erlang runtime system) are written to disk. On some platforms, this function might have no effect.

Typical error reasons are:

enospc

Not enough space left to write the file.

Types:

Truncates the file referenced by IoDevice at the current position. Returns ok if successful, otherwise {error, Reason}.

Types:

Writes Bytes to the file referenced by IoDevice. This function is the only way to write to a file opened in raw mode (although it works for normally opened files, too). Returns ok if successful, and {error, Reason} otherwise.

Typical error reasons are:

ebadf

The file is not opened for writing.

enospc

There is a no space left on the device.

Types:

Writes the contents of the binary data object Binary to the file Filename. The file is created if it does not exist. If it exists, the previous contents are overwritten. Returns ok, or {error, Reason}.

Typical error reasons are:

enoent

A component of the file name does not exist.

enotdir

A component of the file name is not a directory. On some platforms, enoent is returned instead.

enospc

There is a no space left on the device.

eacces

Missing permission for writing the file or searching one of the parent directories.

eisdir

The named file is a directory.

Types:

Same as write_file/2, but takes a third argument Modes, a list of possible modes, see open/2. The mode flags binary and write are implicit, so they should not be used.

Types:

Change file information. Returns ok if successful, otherwise {error, Reason}. FileInfo is a record file_info, defined in the Kernel include file file.hrl. Include the following directive in the module from which the function is called:

-include_lib("kernel/include/file.hrl").
        

The following fields are used from the record, if they are given.

atime = time()

The last (local) time the file was read.

mtime = time()

The last (local) time the file was written.

ctime = time()

On Unix, any value give for this field will be ignored (the "ctime" for the file will be set to the current time). On Windows, this field is the new creation time to set for the file.

mode = int()

The file permissions as the sum of the following bit values:

8#00400

read permission: owner

8#00200

write permission: owner

8#00100

execute permission: owner

8#00040

read permission: group

8#00020

write permission: group

8#00010

execute permission: group

8#00004

read permission: other

8#00002

write permission: other

8#00001

execute permission: other

16#800

set user id on execution

16#400

set group id on execution

On Unix platforms, other bits than those listed above may be set.

uid = int()

Indicates the owner of the file. Ignored for non-Unix file systems.

gid = int()

Gives the group that the owner of the file belongs to. Ignored non-Unix file systems.

Typical error reasons:

eacces

Missing search permission for one of the parent directories of the file.

enoent

The file does not exist.

enotdir

A component of the file name is not a directory. On some platforms, enoent is returned instead.

• eacces - permission denied
• eagain - resource temporarily unavailable
• ebadf - bad file number
• ebusy - file busy
• edquot - disk quota exceeded
• eexist - file already exists
• efault - bad address in system call argument
• efbig - file too large
• eintr - interrupted system call
• einval - invalid argument
• eio - IO error
• eisdir - illegal operation on a directory
• eloop - too many levels of symbolic links
• emfile - too many open files
• emlink - too many links
• enametoolong - file name too long
• enfile - file table overflow
• enodev - no such device
• enoent - no such file or directory
• enomem - not enough memory
• enospc - no space left on device
• enotblk - block device required
• enotdir - not a directory
• enotsup - operation not supported
• enxio - no such device or address
• eperm - not owner
• epipe - broken pipe
• erofs - read-only file system
• espipe - invalid seek
• esrch - no such process
• estale - stale remote file handle
• exdev - cross-domain link

Some operating system file operations, for example a sync/1 or close/1 on a huge file, may block their calling thread for seconds. If this befalls the emulator main thread, the response time is no longer in the order of milliseconds, depending on the definition of "soft" in soft real-time system.

If the device driver thread pool is active, file operations are done through those threads instead, so the emulator can go on executing Erlang processes. Unfortunately, the time for serving a file operation increases due to the extra scheduling required from the operating system.

If the device driver thread pool is disabled or of size 0, large file reads and writes are segmented into several smaller, which enables the emulator so server other processes during the file operation. This gives the same effect as when using the thread pool, but with larger overhead. Other file operations, for example sync/1 or close/1 on a huge file, still are a problem.

For increased performance, raw files are recommended. Raw files uses the file system of the node's host machine. For normal files (non-raw), the file server is used to find the files, and if the node is running its file server as slave to another node's, and the other node runs on some other host machine, they may have different file systems. This is seldom a problem, but you have now been warned.

A normal file is really a process so it can be used as an IO device (see io). Therefore when data is written to a normal file, the sending of the data to the file process, copies all data that are not binaries. Opening the file in binary mode and writing binaries is therefore recommended. If the file is opened on another node, or if the file server runs as slave to another node's, also binaries are copied.

Caching data to reduce the number of file operations, or rather the number of calls to the file driver, will generally increase performance. The following function writes 4 MBytes in 23 seconds when tested:

create_file_slow(Name, N) when integer(N), N >= 0 ->
    {ok, FD} = file:open(Name, [raw, write, delayed_write, binary]),
    ok = create_file_slow(FD, 0, N),
    ok = ?FILE_MODULE:close(FD),
    ok.
      
create_file_slow(FD, M, M) ->
    ok;
create_file_slow(FD, M, N) ->
    ok = file:write(FD, <<M:32/unsigned>>),
    create_file_slow(FD, M+1, N).
    

The following, functionally equivalent, function collects 1024 entries into a list of 128 32-byte binaries before each call to file:write/2 and so does the same work in 0.52 seconds, which is 44 times faster.

create_file(Name, N) when integer(N), N >= 0 ->
    {ok, FD} = file:open(Name, [raw, write, delayed_write, binary]),
    ok = create_file(FD, 0, N),
    ok = ?FILE_MODULE:close(FD),
    ok.
      
create_file(FD, M, M) ->
    ok;
create_file(FD, M, N) when M + 1024 =< N ->
    create_file(FD, M, M + 1024, []),
    create_file(FD, M + 1024, N);
create_file(FD, M, N) ->
    create_file(FD, M, N, []).
      
create_file(FD, M, M, R) ->
    ok = file:write(FD, R);
create_file(FD, M, N0, R) when M + 8 =< N0 ->
    N1  = N0-1,  N2  = N0-2,  N3  = N0-3,  N4  = N0-4, 
    N5  = N0-5,  N6  = N0-6,  N7  = N0-7,  N8  = N0-8, 
    create_file(FD, M, N8, 
                [<<N8:32/unsigned,  N7:32/unsigned, 
                   N6:32/unsigned,  N5:32/unsigned, 
                   N4:32/unsigned,  N3:32/unsigned, 
                   N2:32/unsigned,  N1:32/unsigned>> | R]);
create_file(FD, M, N0, R) ->
    N1 = N0-1,
    create_file(FD, M, N1, [<<N1:32/unsigned>> | R]).
    

Trust only your own benchmarks. If the list length in create_file/2 above is increased, it will run slightly faster, but consume more memory and cause more memory fragmentation. How much this affects your application is something that this simple benchmark can not predict. If the size of each binary is increased to 64 bytes, it will also run slightly faster, but the code will be twice as clumsy. In the current implementation are binaries larger than 64 bytes stored in memory common to all processes and not copied when sent between processes, while these smaller binaries are stored on the process heap and copied when sent like any other term. So, with a binary size of 68 bytes create_file/2 runs 30 percent slower then with 64 bytes, and will cause much more memory fragmentation. Note that if the binaries were to be sent between processes (for example a non-raw file) the results would probably be completely different.

A raw file is really a port. When writing data to a port, it is efficient to write a list of binaries. There is no need to flatten a deep list before writing. On Unix hosts, scatter output, which writes a set of buffers in one operation, is used when possible. In this way file:write(FD, [Bin1, Bin2 | Bin3]) will write the contents of the binaries without copying the data at all except for perhaps deep down in the operating system kernel.

For raw files, pwrite/2 and pread/2 are efficiently implemented. The file driver is called only once for the whole operation, and the list iteration is done in the file driver.

The options delayed_write and read_ahead to file:open/2 makes the file driver cache data to reduce the number of operating system calls. The function create_file/2 in the example above takes 60 seconds seconds without the delayed_write option, which is 2.6 times slower.

And, as a really bad example, create_file_slow/2 above without the raw, binary and delayed_write options, that is it calls file:open(Name, [write]), needs 1 min 20 seconds for the job, which is 3.5 times slower than the first example, and 150 times slower than the optimized create_file/2.

If an error occurs when accessing an open file with the io module, the process which handles the file will exit. The dead file process might hang if a process tries to access it later. This will be fixed in a future release.

filename(3)

Robert Virding - support@erlang.ericsson.se
Claes Wikström - support@erlang.ericsson.se
Björn Gustavsson - support@erlang.ericsson.se