diff options
Diffstat (limited to 'lib/pure/memfiles.nim')
| -rw-r--r-- | lib/pure/memfiles.nim | 306 |
1 files changed, 173 insertions, 133 deletions
diff --git a/lib/pure/memfiles.nim b/lib/pure/memfiles.nim index 277c4ddb6..8eec551c4 100644 --- a/lib/pure/memfiles.nim +++ b/lib/pure/memfiles.nim @@ -16,40 +16,76 @@ ## other "line-like", variable length, delimited records). when defined(windows): - import winlean + import std/winlean + when defined(nimPreviewSlimSystem): + import std/widestrs elif defined(posix): - import posix + import std/posix else: {.error: "the memfiles module is not supported on your operating system!".} -import os, streams +import std/streams +import std/oserrors + +when defined(nimPreviewSlimSystem): + import std/[syncio, assertions] + proc newEIO(msg: string): ref IOError = new(result) result.msg = msg +proc setFileSize(fh: FileHandle, newFileSize = -1, oldSize = -1): OSErrorCode = + ## Set the size of open file pointed to by `fh` to `newFileSize` if != -1, + ## allocating | freeing space from the file system. This routine returns the + ## last OSErrorCode found rather than raising to support old rollback/clean-up + ## code style. [ Should maybe move to std/osfiles. ] + if newFileSize < 0 or newFileSize == oldSize: + return + when defined(windows): + var sizeHigh = int32(newFileSize shr 32) + let sizeLow = int32(newFileSize and 0xffffffff) + let status = setFilePointer(fh, sizeLow, addr(sizeHigh), FILE_BEGIN) + let lastErr = osLastError() + if (status == INVALID_SET_FILE_POINTER and lastErr.int32 != NO_ERROR) or + setEndOfFile(fh) == 0: + result = lastErr + else: + if newFileSize > oldSize: # grow the file + var e: cint # posix_fallocate truncates up when needed. + when declared(posix_fallocate): + while (e = posix_fallocate(fh, 0, newFileSize); e == EINTR): + discard + if e in [EINVAL, EOPNOTSUPP] and ftruncate(fh, newFileSize) == -1: + result = osLastError() # fallback arguable; Most portable BUT allows SEGV + elif e != 0: + result = osLastError() + else: # shrink the file + if ftruncate(fh.cint, newFileSize) == -1: + result = osLastError() + type - MemFile* = object ## represents a memory mapped file - mem*: pointer ## a pointer to the memory mapped file. The pointer - ## can be used directly to change the contents of the - ## file, if it was opened with write access. - size*: int ## size of the memory mapped file + MemFile* = object ## represents a memory mapped file + mem*: pointer ## a pointer to the memory mapped file. The pointer + ## can be used directly to change the contents of the + ## file, if it was opened with write access. + size*: int ## size of the memory mapped file when defined(windows): - fHandle*: Handle ## **Caution**: Windows specific public field to allow - ## even more low level trickery. - mapHandle*: Handle ## **Caution**: Windows specific public field. - wasOpened*: bool ## **Caution**: Windows specific public field. + fHandle*: Handle ## **Caution**: Windows specific public field to allow + ## even more low level trickery. + mapHandle*: Handle ## **Caution**: Windows specific public field. + wasOpened*: bool ## **Caution**: Windows specific public field. else: - handle*: cint ## **Caution**: Posix specific public field. - flags: cint ## **Caution**: Platform specific private field. + handle*: cint ## **Caution**: Posix specific public field. + flags: cint ## **Caution**: Platform specific private field. proc mapMem*(m: var MemFile, mode: FileMode = fmRead, mappedSize = -1, offset = 0, mapFlags = cint(-1)): pointer = ## returns a pointer to a mapped portion of MemFile `m` ## - ## ``mappedSize`` of ``-1`` maps to the whole file, and - ## ``offset`` must be multiples of the PAGE SIZE of your OS + ## `mappedSize` of `-1` maps to the whole file, and + ## `offset` must be multiples of the PAGE SIZE of your OS if mode == fmAppend: raise newEIO("The append mode is not supported.") @@ -60,7 +96,7 @@ proc mapMem*(m: var MemFile, mode: FileMode = fmRead, if readonly: FILE_MAP_READ else: FILE_MAP_READ or FILE_MAP_WRITE, int32(offset shr 32), int32(offset and 0xffffffff), - if mappedSize == -1: 0 else: mappedSize, + WinSizeT(if mappedSize == -1: 0 else: mappedSize), nil) if result == nil: raiseOSError(osLastError()) @@ -83,12 +119,12 @@ proc mapMem*(m: var MemFile, mode: FileMode = fmRead, proc unmapMem*(f: var MemFile, p: pointer, size: int) = - ## unmaps the memory region ``(p, <p+size)`` of the mapped file `f`. + ## unmaps the memory region `(p, <p+size)` of the mapped file `f`. ## All changes are written back to the file system, if `f` was opened ## with write access. ## - ## ``size`` must be of exactly the size that was requested - ## via ``mapMem``. + ## `size` must be of exactly the size that was requested + ## via `mapMem`. when defined(windows): if unmapViewOfFile(p) == 0: raiseOSError(osLastError()) else: @@ -98,27 +134,27 @@ proc unmapMem*(f: var MemFile, p: pointer, size: int) = proc open*(filename: string, mode: FileMode = fmRead, mappedSize = -1, offset = 0, newFileSize = -1, allowRemap = false, mapFlags = cint(-1)): MemFile = - ## opens a memory mapped file. If this fails, ``OSError`` is raised. + ## opens a memory mapped file. If this fails, `OSError` is raised. ## - ## ``newFileSize`` can only be set if the file does not exist and is opened + ## `newFileSize` can only be set if the file does not exist and is opened ## with write access (e.g., with fmReadWrite). ## - ##``mappedSize`` and ``offset`` + ##`mappedSize` and `offset` ## can be used to map only a slice of the file. ## - ## ``offset`` must be multiples of the PAGE SIZE of your OS + ## `offset` must be multiples of the PAGE SIZE of your OS ## (usually 4K or 8K but is unique to your OS) ## - ## ``allowRemap`` only needs to be true if you want to call ``mapMem`` on + ## `allowRemap` only needs to be true if you want to call `mapMem` on ## the resulting MemFile; else file handles are not kept open. ## - ## ``mapFlags`` allows callers to override default choices for memory mapping + ## `mapFlags` allows callers to override default choices for memory mapping ## flags with a bitwise mask of a variety of likely platform-specific flags ## which may be ignored or even cause `open` to fail if misspecified. ## ## Example: ## - ## .. code-block:: nim + ## ```nim ## var ## mm, mm_full, mm_half: MemFile ## @@ -130,6 +166,7 @@ proc open*(filename: string, mode: FileMode = fmRead, ## ## # Read the first 512 bytes ## mm_half = memfiles.open("/tmp/test.mmap", mode = fmReadWrite, mappedSize = 512) + ## ``` # The file can be resized only when write mode is used: if mode == fmAppend: @@ -163,28 +200,17 @@ proc open*(filename: string, mode: FileMode = fmRead, if readonly: shareMode else: shareMode or FILE_SHARE_WRITE, nil, if newFileSize != -1: CREATE_ALWAYS else: OPEN_EXISTING, - if readonly: FILE_ATTRIBUTE_READONLY or flags else: FILE_ATTRIBUTE_NORMAL or flags, + if readonly: FILE_ATTRIBUTE_READONLY or flags + else: FILE_ATTRIBUTE_NORMAL or flags, 0) - when useWinUnicode: - result.fHandle = callCreateFile(createFileW, newWideCString(filename)) - else: - result.fHandle = callCreateFile(createFileA, filename) + result.fHandle = callCreateFile(createFileW, newWideCString(filename)) if result.fHandle == INVALID_HANDLE_VALUE: fail(osLastError(), "error opening file") - if newFileSize != -1: - var - sizeHigh = int32(newFileSize shr 32) - sizeLow = int32(newFileSize and 0xffffffff) - - var status = setFilePointer(result.fHandle, sizeLow, addr(sizeHigh), - FILE_BEGIN) - let lastErr = osLastError() - if (status == INVALID_SET_FILE_POINTER and lastErr.int32 != NO_ERROR) or - (setEndOfFile(result.fHandle) == 0): - fail(lastErr, "error setting file size") + if (let e = setFileSize(result.fHandle.FileHandle, newFileSize); + e != 0.OSErrorCode): fail(e, "error setting file size") # since the strings are always 'nil', we simply always call # CreateFileMappingW which should be slightly faster anyway: @@ -218,7 +244,7 @@ proc open*(filename: string, mode: FileMode = fmRead, result.wasOpened = true if not allowRemap and result.fHandle != INVALID_HANDLE_VALUE: - if closeHandle(result.fHandle) == 0: + if closeHandle(result.fHandle) != 0: result.fHandle = INVALID_HANDLE_VALUE else: @@ -227,48 +253,37 @@ proc open*(filename: string, mode: FileMode = fmRead, if result.handle != -1: discard close(result.handle) raiseOSError(errCode) - var flags = if readonly: O_RDONLY else: O_RDWR + var flags = (if readonly: O_RDONLY else: O_RDWR) or O_CLOEXEC if newFileSize != -1: flags = flags or O_CREAT or O_TRUNC - var permissions_mode = S_IRUSR or S_IWUSR - result.handle = open(filename, flags, permissions_mode) + var permissionsMode = S_IRUSR or S_IWUSR + result.handle = open(filename, flags, permissionsMode) + if result.handle != -1: + if (let e = setFileSize(result.handle.FileHandle, newFileSize); + e != 0.OSErrorCode): fail(e, "error setting file size") else: result.handle = open(filename, flags) if result.handle == -1: - # XXX: errno is supposed to be set here - # Is there an exception that wraps it? fail(osLastError(), "error opening file") - if newFileSize != -1: - if ftruncate(result.handle, newFileSize) == -1: - fail(osLastError(), "error setting file size") - - if mappedSize != -1: - result.size = mappedSize - else: - var stat: Stat + if mappedSize != -1: #XXX Logic here differs from `when windows` branch .. + result.size = mappedSize #.. which always fstats&Uses min(mappedSize, st). + else: # if newFileSize!=-1: result.size=newFileSize # if trust setFileSize + var stat: Stat #^^.. BUT some FSes (eg. Linux HugeTLBfs) round to 2MiB. if fstat(result.handle, stat) != -1: - # XXX: Hmm, this could be unsafe - # Why is mmap taking int anyway? - result.size = int(stat.st_size) + result.size = stat.st_size.int # int may be 32-bit-unsafe for 2..<4 GiB else: fail(osLastError(), "error getting file size") result.flags = if mapFlags == cint(-1): MAP_SHARED else: mapFlags - #Ensure exactly one of MAP_PRIVATE cr MAP_SHARED is set + # Ensure exactly one of MAP_PRIVATE cr MAP_SHARED is set if int(result.flags and MAP_PRIVATE) == 0: result.flags = result.flags or MAP_SHARED - result.mem = mmap( - nil, - result.size, - if readonly: PROT_READ else: PROT_READ or PROT_WRITE, - result.flags, - result.handle, - offset) - + let pr = if readonly: PROT_READ else: PROT_READ or PROT_WRITE + result.mem = mmap(nil, result.size, pr, result.flags, result.handle, offset) if result.mem == cast[pointer](MAP_FAILED): fail(osLastError(), "file mapping failed") @@ -298,34 +313,58 @@ proc flush*(f: var MemFile; attempts: Natural = 3) = if lastErr != EBUSY.OSErrorCode: raiseOSError(lastErr, "error flushing mapping") -when defined(posix) or defined(nimdoc): - proc resize*(f: var MemFile, newFileSize: int) {.raises: [IOError, OSError].} = - ## resize and re-map the file underlying an ``allowRemap MemFile``. - ## **Note**: this assumes the entire file is mapped read-write at offset zero. - ## Also, the value of ``.mem`` will probably change. - ## **Note**: This is not (yet) avaiable on Windows. - when defined(posix): - if f.handle == -1: - raise newException(IOError, - "Cannot resize MemFile opened with allowRemap=false") - if ftruncate(f.handle, newFileSize) == -1: - raiseOSError(osLastError()) - when defined(linux): #Maybe NetBSD, too? - #On Linux this can be over 100 times faster than a munmap,mmap cycle. - proc mremap(old: pointer; oldSize,newSize: csize; flags: cint): pointer {. - importc: "mremap", header: "<sys/mman.h>" .} - let newAddr = mremap(f.mem, csize(f.size), csize(newFileSize), cint(1)) - if newAddr == cast[pointer](MAP_FAILED): - raiseOSError(osLastError()) - else: - if munmap(f.mem, f.size) != 0: - raiseOSError(osLastError()) - let newAddr = mmap(nil, newFileSize, PROT_READ or PROT_WRITE, - f.flags, f.handle, 0) - if newAddr == cast[pointer](MAP_FAILED): - raiseOSError(osLastError()) - f.mem = newAddr +proc resize*(f: var MemFile, newFileSize: int) {.raises: [IOError, OSError].} = + ## Resize & re-map the file underlying an `allowRemap MemFile`. If the OS/FS + ## supports it, file space is reserved to ensure room for new virtual pages. + ## Caller should wait often enough for `flush` to finish to limit use of + ## system RAM for write buffering, perhaps just prior to this call. + ## **Note**: this assumes the entire file is mapped read-write at offset 0. + ## Also, the value of `.mem` will probably change. + if newFileSize < 1: # Q: include system/bitmasks & use PageSize ? + raise newException(IOError, "Cannot resize MemFile to < 1 byte") + when defined(windows): + if not f.wasOpened: + raise newException(IOError, "Cannot resize unopened MemFile") + if f.fHandle == INVALID_HANDLE_VALUE: + raise newException(IOError, + "Cannot resize MemFile opened with allowRemap=false") + if unmapViewOfFile(f.mem) == 0 or closeHandle(f.mapHandle) == 0: # Un-do map + raiseOSError(osLastError()) + if newFileSize != f.size: # Seek to size & `setEndOfFile` => allocated. + if (let e = setFileSize(f.fHandle.FileHandle, newFileSize); + e != 0.OSErrorCode): raiseOSError(e) + f.mapHandle = createFileMappingW(f.fHandle, nil, PAGE_READWRITE, 0,0,nil) + if f.mapHandle == 0: # Re-do map + raiseOSError(osLastError()) + if (let m = mapViewOfFileEx(f.mapHandle, FILE_MAP_READ or FILE_MAP_WRITE, + 0, 0, WinSizeT(newFileSize), nil); m != nil): + f.mem = m f.size = newFileSize + else: + raiseOSError(osLastError()) + elif defined(posix): + if f.handle == -1: + raise newException(IOError, + "Cannot resize MemFile opened with allowRemap=false") + if newFileSize != f.size: + if (let e = setFileSize(f.handle.FileHandle, newFileSize, f.size); + e != 0.OSErrorCode): raiseOSError(e) + when defined(linux): #Maybe NetBSD, too? + # On Linux this can be over 100 times faster than a munmap,mmap cycle. + proc mremap(old: pointer; oldSize, newSize: csize_t; flags: cint): + pointer {.importc: "mremap", header: "<sys/mman.h>".} + let newAddr = mremap(f.mem, csize_t(f.size), csize_t(newFileSize), 1.cint) + if newAddr == cast[pointer](MAP_FAILED): + raiseOSError(osLastError()) + else: + if munmap(f.mem, f.size) != 0: + raiseOSError(osLastError()) + let newAddr = mmap(nil, newFileSize, PROT_READ or PROT_WRITE, + f.flags, f.handle, 0) + if newAddr == cast[pointer](MAP_FAILED): + raiseOSError(osLastError()) + f.mem = newAddr + f.size = newFileSize proc close*(f: var MemFile) = ## closes the memory mapped file `f`. All changes are written back to the @@ -362,7 +401,7 @@ proc close*(f: var MemFile) = if error: raiseOSError(lastErr) -type MemSlice* = object ## represent slice of a MemFile for iteration over delimited lines/records +type MemSlice* = object ## represent slice of a MemFile for iteration over delimited lines/records data*: pointer size*: int @@ -373,10 +412,10 @@ proc `==`*(x, y: MemSlice): bool = proc `$`*(ms: MemSlice): string {.inline.} = ## Return a Nim string built from a MemSlice. result.setLen(ms.size) - copyMem(addr(result[0]), ms.data, ms.size) + copyMem(result.cstring, ms.data, ms.size) -iterator memSlices*(mfile: MemFile, delim='\l', eat='\r'): MemSlice {.inline.} = - ## Iterates over [optional `eat`] `delim`-delimited slices in MemFile `mfile`. +iterator memSlices*(mfile: MemFile, delim = '\l', eat = '\r'): MemSlice {.inline.} = + ## Iterates over \[optional `eat`] `delim`-delimited slices in MemFile `mfile`. ## ## Default parameters parse lines ending in either Unix(\\l) or Windows(\\r\\l) ## style on on a line-by-line basis. I.e., not every line needs the same ending. @@ -399,66 +438,67 @@ iterator memSlices*(mfile: MemFile, delim='\l', eat='\r'): MemSlice {.inline.} = ## functions, not str* functions). ## ## Example: - ## - ## .. code-block:: nim + ## ```nim ## var count = 0 ## for slice in memSlices(memfiles.open("foo")): ## if slice.size > 0 and cast[cstring](slice.data)[0] != '#': ## inc(count) ## echo count + ## ``` - proc c_memchr(cstr: pointer, c: char, n: csize): pointer {. - importc: "memchr", header: "<string.h>" .} + proc c_memchr(cstr: pointer, c: char, n: csize_t): pointer {. + importc: "memchr", header: "<string.h>".} proc `-!`(p, q: pointer): int {.inline.} = return cast[int](p) -% cast[int](q) var ms: MemSlice var ending: pointer ms.data = mfile.mem var remaining = mfile.size while remaining > 0: - ending = c_memchr(ms.data, delim, remaining) - if ending == nil: # unterminated final slice - ms.size = remaining # Weird case..check eat? + ending = c_memchr(ms.data, delim, csize_t(remaining)) + if ending == nil: # unterminated final slice + ms.size = remaining # Weird case..check eat? yield ms break - ms.size = ending -! ms.data # delim is NOT included + ms.size = ending -! ms.data # delim is NOT included if eat != '\0' and ms.size > 0 and cast[cstring](ms.data)[ms.size - 1] == eat: - dec(ms.size) # trim pre-delim char + dec(ms.size) # trim pre-delim char yield ms - ms.data = cast[pointer](cast[int](ending) +% 1) # skip delim + ms.data = cast[pointer](cast[int](ending) +% 1) # skip delim remaining = mfile.size - (ms.data -! mfile.mem) -iterator lines*(mfile: MemFile, buf: var TaintedString, delim='\l', eat='\r'): TaintedString {.inline.} = +iterator lines*(mfile: MemFile, buf: var string, delim = '\l', + eat = '\r'): string {.inline.} = ## Replace contents of passed buffer with each new line, like - ## `readLine(File) <system.html#readLine,File,TaintedString>`_. - ## `delim`, `eat`, and delimiting logic is exactly as for - ## `memSlices <#memSlices>`_, but Nim strings are returned. + ## `readLine(File) <syncio.html#readLine,File,string>`_. + ## `delim`, `eat`, and delimiting logic is exactly as for `memSlices + ## <#memSlices.i,MemFile,char,char>`_, but Nim strings are returned. ## ## Example: - ## - ## .. code-block:: nim - ## var buffer: TaintedString = "" + ## ```nim + ## var buffer: string = "" ## for line in lines(memfiles.open("foo"), buffer): ## echo line + ## ``` for ms in memSlices(mfile, delim, eat): - setLen(buf.string, ms.size) + setLen(buf, ms.size) if ms.size > 0: copyMem(addr buf[0], ms.data, ms.size) yield buf -iterator lines*(mfile: MemFile, delim='\l', eat='\r'): TaintedString {.inline.} = +iterator lines*(mfile: MemFile, delim = '\l', eat = '\r'): string {.inline.} = ## Return each line in a file as a Nim string, like - ## `lines(File) <system.html#lines.i,File>`_. - ## `delim`, `eat`, and delimiting logic is exactly as for - ## `memSlices <#memSlices>`_, but Nim strings are returned. + ## `lines(File) <syncio.html#lines.i,File>`_. + ## `delim`, `eat`, and delimiting logic is exactly as for `memSlices + ## <#memSlices.i,MemFile,char,char>`_, but Nim strings are returned. ## ## Example: - ## - ## .. code-block:: nim + ## ```nim ## for line in lines(memfiles.open("foo")): ## echo line + ## ``` - var buf = TaintedString(newStringOfCap(80)) + var buf = newStringOfCap(80) for line in lines(mfile, buf, delim, eat): yield buf @@ -476,7 +516,7 @@ proc mmsClose(s: Stream) = proc mmsFlush(s: Stream) = flush(MemMapFileStream(s).mf) proc mmsAtEnd(s: Stream): bool = (MemMapFileStream(s).pos >= MemMapFileStream(s).mf.size) or - (MemMapFileStream(s).pos < 0) + (MemMapFileStream(s).pos < 0) proc mmsSetPosition(s: Stream, pos: int) = if pos > MemMapFileStream(s).mf.size or pos < 0: @@ -486,8 +526,8 @@ proc mmsSetPosition(s: Stream, pos: int) = proc mmsGetPosition(s: Stream): int = MemMapFileStream(s).pos proc mmsPeekData(s: Stream, buffer: pointer, bufLen: int): int = - let startAddress = cast[ByteAddress](MemMapFileStream(s).mf.mem) - let p = cast[ByteAddress](MemMapFileStream(s).pos) + let startAddress = cast[int](MemMapFileStream(s).mf.mem) + let p = cast[int](MemMapFileStream(s).pos) let l = min(bufLen, MemMapFileStream(s).mf.size - p) moveMem(buffer, cast[pointer](startAddress + p), l) result = l @@ -502,17 +542,17 @@ proc mmsWriteData(s: Stream, buffer: pointer, bufLen: int) = let size = MemMapFileStream(s).mf.size if MemMapFileStream(s).pos + bufLen > size: raise newEIO("cannot write to stream") - let p = cast[ByteAddress](MemMapFileStream(s).mf.mem) + - cast[ByteAddress](MemMapFileStream(s).pos) + let p = cast[int](MemMapFileStream(s).mf.mem) + + cast[int](MemMapFileStream(s).pos) moveMem(cast[pointer](p), buffer, bufLen) inc(MemMapFileStream(s).pos, bufLen) -proc newMemMapFileStream*(filename: string, mode: FileMode = fmRead, fileSize: int = -1): - MemMapFileStream = +proc newMemMapFileStream*(filename: string, mode: FileMode = fmRead, + fileSize: int = -1): MemMapFileStream = ## creates a new stream from the file named `filename` with the mode `mode`. ## Raises ## `OSError` if the file cannot be opened. See the `system ## <system.html>`_ module for a list of available FileMode enums. - ## ``fileSize`` can only be set if the file does not exist and is opened + ## `fileSize` can only be set if the file does not exist and is opened ## with write access (e.g., with fmReadWrite). var mf: MemFile = open(filename, mode, newFileSize = fileSize) new(result) |