Add a type of DownloadFile that can be split

This allows Kolibri to create content nodes from arbitrarily large
source files for certain types of content.

Author: dylanmccall

ricecooker/classes/files.py

@@ -29,6 +29,7 @@
 from ..exceptions import UnknownFileTypeError
 from ricecooker.utils.encodings import get_base64_encoding
 from ricecooker.utils.encodings import write_base64_to_file
+from ricecooker.utils.file_slice import FileSlice
 from ricecooker.utils.images import create_image_from_epub
 from ricecooker.utils.images import create_image_from_pdf_page
 from ricecooker.utils.images import create_image_from_zip
@@ -114,13 +115,19 @@ def generate_key(action, path_or_id, settings=None, default=" (default)"):
 def get_cache_filename(key):
     cache_file = FILECACHE.get(key)
     if cache_file:
-        cache_file = cache_file.decode("utf-8")
+        cache_file = cache_file.decode("utf-8").split(",")
         # if the file was somehow deleted, make sure we don't return it.
-        if not os.path.exists(config.get_storage_path(cache_file)):
+        if not all(map(cache_file_exists, cache_file)):
             cache_file = None
+    if cache_file and len(cache_file) == 1:
+        cache_file = cache_file[0]
     return cache_file
 
 
+def cache_file_exists(cache_file):
+    return os.path.exists(config.get_storage_path(cache_file))
+
+
 def cache_is_outdated(path, cache_file):
     outdated = True
     if not cache_file:
@@ -137,15 +144,18 @@ def cache_is_outdated(path, cache_file):
     return outdated
 
 
-def download(path, default_ext=None):
+def download(path, default_ext=None, slice_size=None):
     """
     Download `path` and save to storage based on file extension derived from `path`.
     :param path: An URL or a local path
     :param default_ext: fallback ext for file when path does not end with .ext
     :return: filename derived from hash of file contents {md5hash(file)}.ext
     :rtype: sting (path of the form `{md5hash(file at path)}.ext`
     """
-    key = "DOWNLOAD:{}".format(path)
+    if slice_size is not None:
+        key = "DOWNLOAD:{}:{}".format(path, slice_size)
+    else:
+        key = "DOWNLOAD:{}".format(path)
 
     cache_file = get_cache_filename(key)
     if not config.UPDATE and not cache_is_outdated(path, cache_file):
@@ -161,8 +171,12 @@ def download(path, default_ext=None):
         # Get extension of file or use `default_ext` if none found
         if not ext:
             ext = extract_path_ext(path, default_ext=default_ext)
-        filename = copy_file_to_storage(tempf.name, ext=ext)
-        FILECACHE.set(key, bytes(filename, "utf-8"))
+        filename = copy_file_to_storage(tempf.name, ext=ext, slice_size=slice_size)
+        if isinstance(filename, list):
+            cache_value = ",".join(filename)
+        else:
+            cache_value = filename
+        FILECACHE.set(key, bytes(cache_value, "utf-8"))
         config.LOGGER.info("\t--- Downloaded {}".format(filename))
         os.unlink(tempf.name)
 
@@ -242,29 +256,54 @@ def write_path_to_filename(path, write_to_file):
 
 
 def get_hash(filepath):
-    file_hash = hashlib.md5()
     with open(filepath, "rb") as fobj:
-        for chunk in iter(lambda: fobj.read(2097152), b""):
-            file_hash.update(chunk)
+        return get_hash_from_fd(fobj)
+
+
+def get_hash_from_fd(fobj):
+    file_hash = hashlib.md5()
+    for chunk in iter(lambda: fobj.read(2097152), b""):
+        file_hash.update(chunk)
     return file_hash.hexdigest()
 
 
-def copy_file_to_storage(srcfilename, ext=None):
+def copy_file_to_storage(src_file_name, ext=None, slice_size=None):
     """
-    Copy `srcfilename` (filepath) to destination.
+    Copy `src_file_name` (filepath) to destination.
+    If `slice_size` is set, the file will be broken into slices if it exceeds
+    that size in bytes.
     :rtype: None
     """
     if ext is None:
-        ext = extract_path_ext(srcfilename)
+        ext = extract_path_ext(src_file_name)
 
-    hash = get_hash(srcfilename)
-    filename = "{}.{}".format(hash, ext)
-    try:
-        shutil.copy(srcfilename, config.get_storage_path(filename))
-    except shutil.SameFileError:
-        pass
+    filenames = []
 
-    return filename
+    with open(src_file_name, "rb") as src_fd:
+        slices = list(FileSlice.from_file(src_fd, slice_size))
+
+        for slice in slices:
+            slice_hash = get_hash_from_fd(slice)
+            slice.seek(0)
+
+            out_file_name = "{}.{}".format(slice_hash, ext)
+            storage_path = config.get_storage_path(out_file_name)
+
+            try:
+                is_same_file = os.path.samefile(storage_path, src_fd.name)
+            except FileNotFoundError:
+                is_same_file = False
+
+            if not is_same_file:
+                with open(storage_path, "wb") as out_fd:
+                    shutil.copyfileobj(slice, out_fd)
+
+            filenames.append(out_file_name)
+
+    if slice_size is None:
+        return filenames[0]
+    else:
+        return filenames
 
 
 def compress_video_file(filename, ffmpeg_settings):
@@ -490,22 +529,68 @@ def validate(self):
 
     def process_file(self):
         try:
-            self.filename, self.ext = download(self.path, default_ext=self.default_ext)
+            self.filename, self.ext = self._download()
             # don't validate for single-digit extension, or no extension
             if not self.ext:
                 self.ext = extract_path_ext(self.path)
-            return self.filename
         # Catch errors related to reading file path and handle silently
         except HTTP_CAUGHT_EXCEPTIONS as err:
             self.error = str(err)
             config.LOGGER.debug("Failed to download, error is: {}".format(err))
             config.FAILED_FILES.append(self)
             return None
 
+        return self.filename
+
+    def _download(self):
+        return download(self.path, default_ext=self.default_ext)
+
     def __str__(self):
         return self.path
 
 
+class SplittableDownloadFile(DownloadFile):
+    """
+    A type of DownloadFile that will be split into pieces if the source file
+    exceeds `slice_size`. This is separate from DownloadFile because not all
+    content types support file splitting.
+    """
+
+    # 2 GB in bytes
+    slice_size = 2000000000
+
+    def process_file(self):
+        filenames = super(SplittableDownloadFile, self).process_file()
+
+        # TODO: When we call node.add_file, we are assuming files will be
+        #       added in sequence and that order will be maintained. Should we
+        #       add a mechanism where it adds split file order to extra_fields,
+        #       similar to SlideshowNode?
+
+        if isinstance(filenames, list):
+            self.filename = filenames[0]
+            for extra_filename in filenames[1:]:
+                extra_file = self.create_split(extra_filename)
+                self.node.add_file(extra_file)
+
+    def create_split(self, filename):
+        download_file = SplittableDownloadFile(
+            self.path,
+            preset=self.get_preset(),
+            language=self.language,
+            default_ext=self.default_ext,
+            source_url=self.source_url,
+        )
+        download_file.filename = filename
+        download_file.ext = self.ext
+        return download_file
+
+    def _download(self):
+        return download(
+            self.path, default_ext=self.default_ext, slice_size=self.slice_size
+        )
+
+
 IMAGE_EXTENSIONS = {
     file_formats.PNG,
     file_formats.JPG,

ricecooker/classes/nodes.py

@@ -181,9 +181,15 @@ def process_files(self):
         - (optionally) generate thumbnail file from the node's content
         Returns: content-hash based filenames of all the files for this node
         """
-        filenames = []
-        for file in self.files:
-            filenames.append(file.process_file())
+
+        # Items may be added to self.files during file.process_file(), so
+        # we will work with a copy and generate our list of filenames
+        # separately.
+
+        for file in list(self.files):
+            file.process_file()
+
+        filenames = [file.filename for file in self.files]
 
         # Auto-generation of thumbnails happens here if derive_thumbnail or config.THUMBNAILS is set
         if not self.has_thumbnail() and (config.THUMBNAILS or self.derive_thumbnail):

ricecooker/utils/file_slice.py

@@ -0,0 +1,74 @@
+class FileSlice(object):
+    """
+    File-like object that represents a slice of a file, starting from its
+    current offset until `count`. Reads are always relative to the slice's
+    start and end point.
+    """
+
+    def __init__(self, file, count=None):
+        self.file = file
+        self.start = file.tell()
+
+        file.seek(0, 2)
+        self.file_size = file.tell()
+
+        if count is None:
+            count = self.file_size
+
+        count = min(self.file_size - self.start, count)
+        self.end = self.start + count
+
+        # Seek to the end of the file so the next FileSlice object will be
+        # created from that point.
+        file.seek(self.end)
+
+        self.__last_offset = self.start
+
+    @classmethod
+    def from_file(cls, file, chunk_size):
+        slice = cls(file, chunk_size)
+        yield slice
+
+        while slice.end < slice.file_size:
+            slice = cls(file, chunk_size)
+            yield slice
+
+    @property
+    def size(self):
+        return self.end - self.start
+
+    def seek(self, offset, whence=0):
+        if whence == 0:
+            offset = self.start + offset
+        elif whence == 1:
+            offset = self.tell() + offset
+        elif whence == 2:
+            offset = self.end - offset
+        self.file.seek(offset)
+        self.__store_offset()
+
+    def __reset_offset(self):
+        if self.file.tell() != self.__last_offset:
+            self.file.seek(self.__last_offset)
+
+    def __store_offset(self):
+        self.__last_offset = self.file.tell()
+
+    def tell(self):
+        self.__reset_offset()
+        return self.file.tell() - self.start
+
+    def read(self, count=None):
+        self.__reset_offset()
+
+        if count is None:
+            count = self.size
+
+        remaining = max(0, self.size - self.tell())
+
+        buffer = self.file.read(min(count, remaining))
+        self.__store_offset()
+        return buffer
+
+    def write(self, string):
+        raise NotImplementedError()