appypod-rattail/pod/doc_importers.py

337 lines
15 KiB
Python
Raw Normal View History

2009-06-29 07:06:01 -05:00
# ------------------------------------------------------------------------------
# Appy is a framework for building applications in the Python language.
# Copyright (C) 2007 Gaetan Delannay
# This program is free software; you can redistribute it and/or
# modify it under the terms of the GNU General Public License
# as published by the Free Software Foundation; either version 2
# of the License, or (at your option) any later version.
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
# GNU General Public License for more details.
# You should have received a copy of the GNU General Public License
# along with this program; if not, write to the Free Software
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301,USA.
# ------------------------------------------------------------------------------
import os, os.path, time, shutil, struct, random, urlparse
2009-06-29 07:06:01 -05:00
from appy.pod import PodError
from appy.pod.odf_parser import OdfEnvironment
from appy.shared import mimeTypesExts
from appy.shared.utils import FileWrapper
from appy.shared.dav import Resource
2009-06-29 07:06:01 -05:00
# ------------------------------------------------------------------------------
FILE_NOT_FOUND = "'%s' does not exist or is not a file."
PDF_TO_IMG_ERROR = 'A PDF file could not be converted into images. Please ' \
'ensure that Ghostscript (gs) is installed on your ' \
'system and the "gs" program is in the path.'
CONVERT_ERROR = 'Program "convert", from imagemagick, must be installed and ' \
'in the path for converting a SVG file into a PNG file.'
2009-06-29 07:06:01 -05:00
# ------------------------------------------------------------------------------
class DocImporter:
'''Base class used for importing external content into a pod template (an
image, another pod template, another odt document...'''
def __init__(self, content, at, format, renderer):
2009-06-29 07:06:01 -05:00
self.content = content
# If content is None, p_at tells us where to find it (file system path,
# url, etc)
self.at = at
# Ensure this path exists, if it is a local path.
if at and not at.startswith('http') and not os.path.isfile(at):
raise PodError(FILE_NOT_FOUND % at)
2009-06-29 07:06:01 -05:00
self.format = format
self.res = u''
self.renderer = renderer
self.ns = renderer.currentParser.env.namespaces
2009-06-29 07:06:01 -05:00
# Unpack some useful namespaces
self.textNs = self.ns[OdfEnvironment.NS_TEXT]
self.linkNs = self.ns[OdfEnvironment.NS_XLINK]
self.drawNs = self.ns[OdfEnvironment.NS_DRAW]
self.svgNs = self.ns[OdfEnvironment.NS_SVG]
self.tempFolder = renderer.tempFolder
2009-06-29 07:06:01 -05:00
self.importFolder = self.getImportFolder()
# Create the import folder if it does not exist.
if not os.path.exists(self.importFolder): os.mkdir(self.importFolder)
self.importPath = self.getImportPath(at, format)
# A link to the global fileNames dict (explained in renderer.py)
self.fileNames = renderer.fileNames
if at:
# Move the file within the ODT, if it is an image and if this image
# has not already been imported.
self.importPath = self.moveFile(at, self.importPath)
2009-06-29 07:06:01 -05:00
else:
# We need to dump the file content (in self.content) in a temp file
# first. self.content may be binary, a file handler or a
# FileWrapper.
2009-06-29 07:06:01 -05:00
if isinstance(self.content, file):
fileContent = self.content.read()
elif isinstance(self.content, FileWrapper):
fileContent = content.content
2009-06-29 07:06:01 -05:00
else:
fileContent = self.content
f = file(self.importPath, 'wb')
f.write(fileContent)
f.close()
# ImageImporter adds additional, image-specific attrs, through
# ImageImporter.setImageInfo.
2009-06-29 07:06:01 -05:00
def getImportFolder(self):
'''This method must be overridden and gives the path where to dump the
content of the document or image. In the case of a document it is a
temp folder; in the case of an image it is a folder within the ODT
result.'''
def getImportPath(self, at, format):
'''Gets the path name of the file to dump on disk (within the ODT for
images, in a temp folder for docs).'''
if not format:
if at.startswith('http'):
format = '' # We will know it only after the HTTP GET.
else:
format = os.path.splitext(at)[1][1:]
fileName = 'f.%d.%f.%s' % (random.randint(0,10), time.time(), format)
return os.path.abspath('%s/%s' % (self.importFolder, fileName))
def moveFile(self, at, importPath):
2009-06-29 07:06:01 -05:00
'''In the case parameter "at" was used, we may want to move the file at
p_at within the ODT result in p_importPath (for images) or do
nothing (for docs). In the latter case, the file to import stays
at _at, and is not copied into p_importPath.'''
2009-06-29 07:06:01 -05:00
return at
class OdtImporter(DocImporter):
'''This class allows to import the content of another ODT document into a
pod template.'''
def getImportFolder(self): return '%s/docImports' % self.tempFolder
2009-06-29 07:06:01 -05:00
def run(self):
self.res += '<%s:section %s:name="PodImportSection%f">' \
'<%s:section-source %s:href="%s" ' \
'%s:filter-name="writer8"/></%s:section>' % (
self.textNs, self.textNs, time.time(), self.textNs,
self.linkNs, self.importPath, self.textNs, self.textNs)
return self.res
class PdfImporter(DocImporter):
'''This class allows to import the content of a PDF file into a pod
template. It calls gs to split the PDF into images and calls the
ImageImporter for importing it into the result.'''
imagePrefix = 'PdfPart'
def getImportFolder(self): return '%s/docImports' % self.tempFolder
2009-06-29 07:06:01 -05:00
def run(self):
# Split the PDF into images with Ghostscript
imagesFolder = os.path.dirname(self.importPath)
cmd = 'gs -dNOPAUSE -dBATCH -sDEVICE=jpeg -r125x125 ' \
'-sOutputFile=%s/%s%%d.jpg %s' % \
(imagesFolder, self.imagePrefix, self.importPath)
os.system(cmd)
# Check that at least one image was generated
succeeded = False
firstImage = '%s1.jpg' % self.imagePrefix
for fileName in os.listdir(imagesFolder):
if fileName == firstImage:
succeeded = True
break
if not succeeded: raise PodError(PDF_TO_IMG_ERROR)
2009-06-29 07:06:01 -05:00
# Insert images into the result.
noMoreImages = False
i = 0
while not noMoreImages:
i += 1
nextImage = '%s/%s%d.jpg' % (imagesFolder, self.imagePrefix, i)
if os.path.exists(nextImage):
# Use internally an Image importer for doing this job.
imgImporter =ImageImporter(None, nextImage, 'jpg',self.renderer)
imgImporter.setImageInfo('paragraph', True, None, None, None)
2009-06-29 07:06:01 -05:00
self.res += imgImporter.run()
os.remove(nextImage)
else:
noMoreImages = True
return self.res
# Compute size of images -------------------------------------------------------
jpgTypes = ('jpg', 'jpeg')
pxToCm = 44.173513561
def getSize(filePath, fileType):
'''Gets the size of an image by reading first bytes.'''
x, y = (None, None)
f = file(filePath, 'rb')
2009-06-29 07:06:01 -05:00
if fileType in jpgTypes:
# Dummy read to skip header ID
f.read(2)
while True:
# Extract the segment header.
(marker, code, length) = struct.unpack("!BBH", f.read(4))
# Verify that it's a valid segment.
if marker != 0xFF:
# No JPEG marker
break
elif code >= 0xC0 and code <= 0xC3:
# Segments that contain size info
(y, x) = struct.unpack("!xHH", f.read(5))
break
else:
# Dummy read to skip over data
f.read(length-2)
elif fileType == 'png':
# Dummy read to skip header data
f.read(12)
if f.read(4) == "IHDR":
x, y = struct.unpack("!LL", f.read(8))
elif fileType == 'gif':
imgType = f.read(6)
buf = f.read(5)
if len(buf) == 5:
# else: invalid/corrupted GIF (bad header)
x, y, u = struct.unpack("<HHB", buf)
f.close()
if x and y:
return float(x)/pxToCm, float(y)/pxToCm
else:
return x, y
2009-06-29 07:06:01 -05:00
class ImageImporter(DocImporter):
'''This class allows to import into the ODT result an image stored
externally.'''
anchorTypes = ('page', 'paragraph', 'char', 'as-char')
WRONG_ANCHOR = 'Wrong anchor. Valid values for anchors are: %s.'
pictFolder = '%sPictures%s' % (os.sep, os.sep)
def getImportFolder(self):
return os.path.join(self.tempFolder, 'unzip', 'Pictures')
def moveFile(self, at, importPath):
'''Copies file at p_at into the ODT file at p_importPath.'''
# Has this image already been imported ?
for imagePath, imageAt in self.fileNames.iteritems():
if imageAt == at:
# Yes!
i = importPath.rfind(self.pictFolder) + 1
return importPath[:i] + imagePath
# The image has not already been imported: copy it.
if not at.startswith('http'):
shutil.copy(at, importPath)
return importPath
# The image must be retrieved via a URL. Try to perform a HTTP GET.
response = Resource(at).get()
if response.code == 200:
# At last, I can get the file format.
self.format = mimeTypesExts[response.headers['Content-Type']]
importPath += self.format
f = file(importPath, 'wb')
f.write(response.body)
f.close()
return importPath
# The HTTP GET did not work, maybe for security reasons (we probably
# have no permission to get the file). But maybe the URL was a local
# one, from an application server running this POD code. In this case,
# if an image resolver has been given to POD, use it to retrieve the
# image.
imageResolver = self.renderer.imageResolver
if not imageResolver:
# Return some default image explaining that the image wasn't found.
import appy.pod
podFolder = os.path.dirname(appy.pod.__file__)
img = os.path.join(podFolder, 'imageNotFound.jpg')
self.format = 'jpg'
importPath += self.format
f = file(img)
imageContent = f.read()
f.close()
f = file(importPath, 'wb')
f.write(imageContent)
f.close()
else:
# The imageResolver is a Zope application. From it, we will
# retrieve the object on which the image is stored and get
# the file to download.
urlParts = urlparse.urlsplit(at)
path = urlParts[2][1:]
obj = imageResolver.unrestrictedTraverse(path.split('/')[:-1])
zopeFile = getattr(obj, urlParts[3].split('=')[1])
appyFile = FileWrapper(zopeFile)
self.format = mimeTypesExts[appyFile.mimeType]
importPath += self.format
appyFile.dump(importPath)
return importPath
def setImageInfo(self, anchor, wrapInPara, size, sizeUnit, style):
# Initialise anchor
2009-06-29 07:06:01 -05:00
if anchor not in self.anchorTypes:
raise PodError(self.WRONG_ANCHOR % str(self.anchorTypes))
self.anchor = anchor
self.wrapInPara = wrapInPara
self.size = size
self.sizeUnit = sizeUnit
# Put CSS attributes from p_style in a dict.
self.cssAttrs = {}
if style:
for attr in style.split(';'):
if not attr.strip(): continue
name, value = attr.strip().split(':')
value = value.strip()
if value.endswith('px'): value = value[:-2]
if value.isdigit(): value=int(value)
self.cssAttrs[name.strip()] = value
2009-06-29 07:06:01 -05:00
def run(self):
# Some shorcuts for the used xml namespaces
d = self.drawNs
t = self.textNs
x = self.linkNs
s = self.svgNs
imageName = 'Image%f' % time.time()
# Compute path to image
i = self.importPath.rfind(self.pictFolder)
imagePath = self.importPath[i+1:].replace('\\', '/')
self.fileNames[imagePath] = self.at
# In the case of SVG files, perform an image conversion to PNG
if imagePath.endswith('.svg'):
newImportPath = os.path.splitext(self.importPath)[0] + '.png'
err= os.system('convert "%s" "%s"'% (self.importPath,newImportPath))
if err:
raise Exception(CONVERT_ERROR)
os.remove(self.importPath)
self.importPath = newImportPath
imagePath = os.path.splitext(imagePath)[0] + '.png'
self.format = 'png'
# Retrieve image size from self.size.
width = height = None
if self.size:
width, height = self.size
if self.sizeUnit == 'px':
# Convert it to cm
width = float(width) / pxToCm
height = float(height) / pxToCm
# Override self.size if 'height' or 'width' is found in self.cssAttrs
if 'width' in self.cssAttrs:
width = float(self.cssAttrs['width']) / pxToCm
if 'height' in self.cssAttrs:
height = float(self.cssAttrs['height']) / pxToCm
# If width and/or height is missing, compute it.
if not width or not height:
width, height = getSize(self.importPath, self.format)
2009-06-29 07:06:01 -05:00
if width != None:
size = ' %s:width="%fcm" %s:height="%fcm"' % (s, width, s, height)
else:
size = ''
if 'float' in self.cssAttrs:
floatValue = self.cssAttrs['float'].capitalize()
styleInfo = '%s:style-name="podImage%s" ' % (d, floatValue)
self.anchor = 'char'
else:
styleInfo = ''
image = '<%s:frame %s%s:name="%s" %s:z-index="0" ' \
'%s:anchor-type="%s"%s><%s:image %s:type="simple" ' \
'%s:show="embed" %s:href="%s" %s:actuate="onLoad"/>' \
'</%s:frame>' % (d, styleInfo, d, imageName, d, t, self.anchor,
size, d, x, x, x, imagePath, x, d)
if hasattr(self, 'wrapInPara') and self.wrapInPara:
image = '<%s:p>%s</%s:p>' % (t, image, t)
self.res += image
2009-06-29 07:06:01 -05:00
return self.res
# ------------------------------------------------------------------------------