Design¶

Per application, there is MemoryManager which is held as static instance and used throughout the application. It can be configured to keep your resources within certain limits.

To access mapped regions, you require a cursor. Cursors point to exactly one file and serve as handles into it. As long as it exists, the respective memory region will remain available.

For convenience, a buffer implementation is provided which handles cursors and resource allocation behind its simple buffer like interface.

Memory Managers¶

There are two types of memory managers, one uses static windows, the other one uses sliding windows. A window is a region of a file mapped into memory. Although the names might be somewhat misleading as technically windows are always static, the sliding version will allocate relatively small windows whereas the static version will always map the whole file.

The static manager does nothing more than keeping a client count on the respective memory maps which always map the whole file, which allows to make some assumptions that can lead to simplified data access and increased performance, but reduces the compatibility to 32 bit systems or giant files.

The sliding memory manager therefore should be the default manager when preparing an application for handling huge amounts of data on 32 bit and 64 bit platforms:

import smmap
# This instance should be globally available in your application
# It is configured to be well suitable for 32-bit or 64 bit applications.
mman = smmap.SlidingWindowMapManager()

# the manager provides much useful information about its current state
# like the amount of open file handles or the amount of mapped memory
mman.num_file_handles()
mman.mapped_memory_size()
# and many more ...

import smmap.test.lib
fc = smmap.test.lib.FileCreator(1024*1024*8, "test_file")

# obtain a cursor to access some file.
c = mman.make_cursor(fc.path)

# the cursor is now associated with the file, but not yet usable
assert c.is_associated()
assert not c.is_valid()

# before you can use the cursor, you have to specify a window you want to
# access. The following just says you want as much data as possible starting
# from offset 0.
# To be sure your region could be mapped, query for validity
assert c.use_region().is_valid()            # use_region returns self

# once a region was mapped, you must query its dimension regularly
# to assure you don't try to access its buffer out of its bounds
assert c.size()
c.buffer()[0]                       # first byte
c.buffer()[1:10]                    # first 9 bytes
c.buffer()[c.size()-1]      # last byte

# its recommended not to create big slices when feeding the buffer
# into consumers (e.g. struct or zlib).
# Instead, either give the buffer directly, or use pythons buffer command.
buffer(c.buffer(), 1, 9)    # first 9 bytes without copying them

# you can query absolute offsets, and check whether an offset is included
# in the cursor's data.
assert c.ofs_begin() < c.ofs_end()
assert c.includes_ofs(100)

# If you are over out of bounds with one of your region requests, the
# cursor will be come invalid. It cannot be used in that state
assert not c.use_region(fc.size, 100).is_valid()
# map as much as possible after skipping the first 100 bytes
assert c.use_region(100).is_valid()

# You can explicitly free cursor resources by unusing the cursor's region
c.unuse_region()
assert not c.is_valid()

Buffers¶

To make first use easier, at the expense of performance, there is a Buffer implementation which uses a cursor underneath.

With it, you can access all data in a possibly huge file without having to take care of setting the cursor to different regions yourself:

# Create a default buffer which can operate on the whole file
buf = smmap.SlidingWindowMapBuffer(mman.make_cursor(fc.path))

# you can use it right away
assert buf.cursor().is_valid()

buf[0]      # access the first byte
buf[-1]     # access the last ten bytes on the file
buf[-10:]# access the last ten bytes

# If you want to keep the instance between different accesses, use the
# dedicated methods
buf.end_access()
assert not buf.cursor().is_valid()  # you cannot use the buffer anymore
assert buf.begin_access(offset=10)  # start using the buffer at an offset

# it will stop using resources automatically once it goes out of scope

Mapped Memory Managers¶

Module containing a memory memory manager which provides a sliding window on a number of memory mapped files

class smmap.mman.StaticWindowMapManager(window_size=0, max_memory_size=0, max_open_handles=9223372036854775807)¶

Provides a manager which will produce single size cursors that are allowed to always map the whole file.

Clients must be written to specifically know that they are accessing their data through a StaticWindowMapManager, as they otherwise have to deal with their window size.

These clients would have to use a SlidingWindowMapBuffer to hide this fact.

This type will always use a maximum window size, and optimize certain methods to accommodate this fact

MapRegionCls¶

alias of MapRegion

MapRegionListCls¶

alias of MapRegionList

MapWindowCls¶

alias of MapWindow

WindowCursorCls¶

alias of WindowCursor

collect()¶

Collect all available free-to-collect mapped regions :return: Amount of freed handles

force_map_handle_removal_win(base_path)¶

ONLY AVAILABLE ON WINDOWS On windows removing files is not allowed if anybody still has it opened. If this process is ourselves, and if the whole process uses this memory manager (as far as the parent framework is concerned) we can enforce closing all memory maps whose path matches the given base path to allow the respective operation after all. The respective system must NOT access the closed memory regions anymore ! This really may only be used if you know that the items which keep the cursors alive will not be using it anymore. They need to be recreated ! :return: Amount of closed handles

Note: does nothing on non-windows platforms

make_cursor(path_or_fd)¶

Returns:	a cursor pointing to the given path or file descriptor. It can be used to map new regions of the file into memory

Note: if a file descriptor is given, it is assumed to be open and valid, but may be closed afterwards. To refer to the same file, you may reuse your existing file descriptor, but keep in mind that new windows can only be mapped as long as it stays valid. This is why the using actual file paths are preferred unless you plan to keep the file descriptor open.

Note: file descriptors are problematic as they are not necessarily unique, as two different files opened and closed in succession might have the same file descriptor id.

Note: Using file descriptors directly is faster once new windows are mapped as it prevents the file to be opened again just for the purpose of mapping it.

mapped_memory_size()¶

Returns:	amount of bytes currently mapped in total

max_file_handles()¶

Returns:	maximium amount of handles we may have opened

max_mapped_memory_size()¶

Returns:	maximum amount of memory we may allocate

num_file_handles()¶

Returns:	amount of file handles in use. Each mapped region uses one file handle

num_open_files()¶

Amount of opened files in the system

window_size()¶

Returns:	size of each window when allocating new regions

class smmap.mman.SlidingWindowMapManager(window_size=-1, max_memory_size=0, max_open_handles=9223372036854775807)¶

Maintains a list of ranges of mapped memory regions in one or more files and allows to easily obtain additional regions assuring there is no overlap. Once a certain memory limit is reached globally, or if there cannot be more open file handles which result from each mmap call, the least recently used, and currently unused mapped regions are unloaded automatically.

Note: currently not thread-safe !

Note: in the current implementation, we will automatically unload windows if we either cannot

create more memory maps (as the open file handles limit is hit) or if we have allocated more than a safe amount of memory already, which would possibly cause memory allocations to fail as our address space is full.

class smmap.mman.WindowCursor(manager=None, regions=None)¶

Pointer into the mapped region of the memory manager, keeping the map alive until it is destroyed and no other client uses it.

Cursors should not be created manually, but are instead returned by the SlidingWindowMapManager

Note:: The current implementation is suited for static and sliding window managers, but it also means that it must be suited for the somewhat quite different sliding manager. It could be improved, but I see no real need to do so.

assign(rhs)¶

Assign rhs to this instance. This is required in order to get a real copy. Alternativly, you can copy an existing instance using the copy module

buffer()¶

Return a buffer object which allows access to our memory region from our offset to the window size. Please note that it might be smaller than you requested when calling use_region()

Note: You can only obtain a buffer if this instance is_valid() !

Note: buffers should not be cached passed the duration of your access as it will prevent resources from being freed even though they might not be accounted for anymore !

fd()¶

Returns:	file descriptor used to create the underlying mapping.

Note: it is not required to be valid anymore :raise ValueError: if the mapping was not created by a file descriptor

file_size()¶

Returns:	size of the underlying file

includes_ofs(ofs)¶

Returns:	True if the given absolute offset is contained in the cursors current region

Note: cursor must be valid for this to work

is_associated()¶

Returns:	True if we are associated with a specific file already

is_valid()¶

Returns:	True if we have a valid and usable region

map()¶

Returns:	the underlying raw memory map. Please not that the offset and size is likely to be different to what you set as offset and size. Use it only if you are sure about the region it maps, which is the whole file in case of StaticWindowMapManager

ofs_begin()¶

Returns:	offset to the first byte pointed to by our cursor

Note: only if is_valid() is True

ofs_end()¶

Returns:	offset to one past the last available byte

path()¶

Returns:	path of the underlying mapped file
Raises ValueError:
	if attached path is not a path

path_or_fd()¶

Returns:	path or file descriptor of the underlying mapped file

region_ref()¶

Returns:	weak ref to our mapped region.
Raises AssertionError:
	if we have no current region. This is only useful for debugging

size()¶

Returns:	amount of bytes we point to

unuse_region()¶

Unuse the ucrrent region. Does nothing if we have no current region

Note: the cursor unuses the region automatically upon destruction. It is recommended to un-use the region once you are done reading from it in persistent cursors as it helps to free up resource more quickly

use_region(offset=0, size=0, flags=0)¶

Assure we point to a window which allows access to the given offset into the file

Parameters:	offset – absolute offset in bytes into the file size – amount of bytes to map. If 0, all available bytes will be mapped flags – additional flags to be given to os.open in case a file handle is initially opened for mapping. Has no effect if a region can actually be reused.
Returns:	this instance - it should be queried for whether it points to a valid memory region. This is not the case if the mapping failed because we reached the end of the file

Note:: The size actually mapped may be smaller than the given size. If that is the case, either the file has reached its end, or the map was created between two existing regions

Buffers¶

Module with a simple buffer implementation using the memory manager

class smmap.buf.SlidingWindowMapBuffer(cursor=None, offset=0, size=9223372036854775807, flags=0)¶

A buffer like object which allows direct byte-wise object and slicing into memory of a mapped file. The mapping is controlled by the provided cursor.

The buffer is relative, that is if you map an offset, index 0 will map to the first byte at the offset you used during initialization or begin_access

Note: Although this type effectively hides the fact that there are mapped windows underneath, it can unfortunately not be used in any non-pure python method which needs a buffer or string

begin_access(cursor=None, offset=0, size=9223372036854775807, flags=0)¶

Call this before the first use of this instance. The method was already called by the constructor in case sufficient information was provided.

For more information no the parameters, see the __init__ method :param path: if cursor is None the existing one will be used. :return: True if the buffer can be used

cursor()¶

Returns:	the currently set cursor which provides access to the data

end_access()¶

Call this method once you are done using the instance. It is automatically called on destruction, and should be called just in time to allow system resources to be freed.

Once you called end_access, you must call begin access before reusing this instance!

Exceptions¶

Module with system exceptions

exception smmap.exc.MemoryManagerError¶

Base class for all exceptions thrown by the memory manager

exception smmap.exc.RegionCollectionError¶

Thrown if a memory region could not be collected, or if no region for collection was found

Utilities¶

Module containing a memory memory manager which provides a sliding window on a number of memory mapped files

smmap.util.align_to_mmap(num, round_up)¶

Align the given integer number to the closest page offset, which usually is 4096 bytes.

Parameters:	round_up – if True, the next higher multiple of page size is used, otherwise the lower page_size will be used (i.e. if True, 1 becomes 4096, otherwise it becomes 0)
Returns:	num rounded to closest page

smmap.util.is_64_bit()¶

Returns:	True if the system is 64 bit. Otherwise it can be assumed to be 32 bit

class smmap.util.buffer¶

buffer(object [, offset[, size]])

Create a new buffer object which references the given object. The buffer will reference a slice of the target object from the start of the object (or at the specified offset). The slice will extend to the end of the target object (or with the specified size).

class smmap.util.MapWindow(offset, size)¶

Utility type which is used to snap windows towards each other, and to adjust their size

align()¶

Assures the previous window area is contained in the new one

extend_left_to(window, max_size)¶

Adjust the offset to start where the given window on our left ends if possible, but don’t make yourself larger than max_size. The resize will assure that the new window still contains the old window area

extend_right_to(window, max_size)¶

Adjust the size to make our window end where the right window begins, but don’t get larger than max_size

classmethod from_region(region)¶

Returns:	new window from a region

ofs¶

ofs_end()¶

size¶

class smmap.util.MapRegion(path_or_fd, ofs, size, flags=0)¶

Defines a mapped region of memory, aligned to pagesizes

Note: deallocates used region automatically on destruction

buffer()¶

Returns:	a buffer containing the memory

client_count()¶

Returns:	number of clients currently using this region

includes_ofs(ofs)¶

Returns:	True if the given offset can be read in our mapped region

increment_usage_count()¶

Adjust the usage count by the given positive or negative offset

map()¶

Returns:	a memory map containing the memory

ofs_begin()¶

Returns:	absolute byte offset to the first byte of the mapping

ofs_end()¶

Returns:	Absolute offset to one byte beyond the mapping into the file

size()¶

Returns:	total size of the mapped region in bytes

usage_count()¶

Returns:	amount of usages so far

class smmap.util.MapRegionList(path_or_fd)¶

List of MapRegion instances associating a path with a list of regions.

client_count()¶

Returns:	amount of clients which hold a reference to this instance

file_size()¶

Returns:	size of file we manager

path_or_fd()¶

Returns:	path or file descriptor we are attached to

v0.8.3¶

• Cleaned up code and assured it works sufficiently well with python 3

v0.8.1¶

• A single bugfix

v0.8.0¶

• Initial Release