import sys import os import re import itertools import tempfile import locale from collections import namedtuple from PyQt5.QtWidgets import QMessageBox from pdfminer.pdfinterp import PDFResourceManager , PDFPageInterpreter from pdfminer.converter import PDFPageAggregator from pdfminer.layout import LAParams , LTTextBoxHorizontal from pdfminer.pdfpage import PDFPage import ghostscript from PIL import Image , ImageChops from asl_cards.db import AslCard , AslCardImage # --------------------------------------------------------------------- class PdfParser: def __init__( self , index_dir , ask=None , progress=None , progress2=None ) : # initialize self.index_dir = index_dir self.ask = ask # nb: for asking the user something during processing self.progress = progress # nb: for tracking file progress self.progress2 = progress2 # nb: for tracking page progress within a file self.cancelling = False def parse( self , target , max_pages=-1 , images=True ) : """Extract the cards from a PDF file.""" # locate the files we're going to parse if os.path.isfile( target ) : fnames = [ target ] else : fnames = [ os.path.join( target , f ) for f in os.listdir( target ) if os.path.splitext( f )[1].lower() == ".pdf" ] # parse each file cards = [] for file_no,fname in enumerate(fnames) : if self.cancelling : raise RuntimeError("Cancelled.") file_cards = self._do_parse_file( float(file_no)/len(fnames) , fname , max_pages , images ) if file_cards : cards.extend( file_cards ) self._progress( 1.0 , "Done." ) # filter out placeholder cards cards = [ c for c in cards if c.nationality != "_unused_" and c.name != "_unused_" ] return cards def _do_parse_file( self , pval , fname , max_pages , images ) : cards = [] # check if we have an index for this file # NOTE: We originally tried to get the details of each card by parsing the PDF files but unfortunately, # the text was coming out garbled. We allow corrections to be supplied in an external file, but if we're # going to do that, we might as well not bother parsing the PDF :-/ (especially since it's so insanely slow). split = os.path.split( fname ) index_fname = os.path.join( self.index_dir if self.index_dir else "" , os.path.splitext(split[1])[0]+".txt" ) if os.path.isfile( index_fname ) : # yup - just generate the AslCard's from that # NOTE: It would be nice to store these files as JSON, or something similar, but we want # to keep them easy for end-users to change, if some values need to be tweaked. self._progress( pval , "Reading card details from {}...".format( os.path.split(index_fname)[1] ) ) for line_buf in open(index_fname,"r") : line_buf = line_buf.strip() if line_buf == "" or line_buf.startswith(("#","'",";","//")) : continue fields = line_buf.split( "|" ) if len(fields) != 3 : raise RuntimeError( "Invalid index line: {}".format( line_buf ) ) fields = [ f.strip() for f in fields ] ncards = len( cards ) cards.append( AslCard( card_tag = fields[0] , nationality = fields[1] , name = fields[2] , page_id = 1 + ncards/2 , page_pos = ncards % 2 ) ) else : # ask the user if they want to try parsing the PDF if self.ask : rc = self.ask( "Can't find an index file for {}.\n\nDo you want to try parsing the PDF (slow and unreliable)?".format( os.path.split( fname )[ 1 ] ) , QMessageBox.Yes | QMessageBox.No , QMessageBox.No ) if rc != QMessageBox.Yes : return None # extract each AslCard from the file self._progress( pval , "Analyzing {}...".format( os.path.split(fname)[1] ) ) rmgr = PDFResourceManager() laparams = LAParams() dev = PDFPageAggregator( rmgr , laparams=laparams ) interp = PDFPageInterpreter( rmgr , dev ) with open(fname,"rb") as fp : pages = list( PDFPage.get_pages( fp ) ) for page_no,page in enumerate(pages) : if self.cancelling : raise RuntimeError("Cancelled.") self._progress2( float(page_no) / len(pages) ) page_cards = self._parse_page( cards , interp , page_no , page ) cards.extend( page_cards ) if max_pages > 0 and 1+page_no >= max_pages : break # extract the card images if images : self._progress( pval , "Extracting images from {}...".format( os.path.split(fname)[1] ) ) card_images = self._extract_images( fname , max_pages ) if len(cards) != len(card_images) : raise RuntimeError( "Card mismatch in {}: found {} cards, {} card images.".format( fname , len(cards) , len(card_images) ) ) for i in range(0,len(cards)) : if self.cancelling : raise RuntimeError("Cancelled.") cards[i].card_image = AslCardImage( image_data=card_images[i] ) return cards def _parse_page( self , cards , interp , page_no , page ) : """Extract the cards from a PDF page.""" cards = [] interp.process_page( page ) lt_page = interp.device.get_result() # locate the info box for each card (in the top-left corner) info_boxes = [] for item in lt_page : if self.cancelling : raise RuntimeError("Cancelled.") if type(item) is not LTTextBoxHorizontal : continue item_text = item.get_text().strip() if item_text.startswith( ("Vehicle","Ordnance") ) : info_boxes.append( [item] ) # get the details from each info box for item in lt_page : if self.cancelling : raise RuntimeError("Cancelled.") if type(item) is not LTTextBoxHorizontal : continue # check if the next item could be part of an info box - it must be within the left/right boundary # of the first item (within a certain tolerance), and below it (but not too far) eps = 50 # left/right tolerance for info_box in info_boxes : if item.x0 >= info_box[0].x0 - eps and item.x1 <= info_box[0].x1 + eps \ and item.y1 < info_box[0].y1 and info_box[0].y0 - item.y1 < 50 : # yup - save it info_box.append( item ) # generate an AslCard from each info box for info_box in info_boxes : card = self._make_asl_card( lt_page , info_box ) cards.append( card ) return cards def _make_asl_card( self , lt_page , items ) : # sort the items vertically items.sort( key=lambda i: i.y0 , reverse=True ) # split out each line of item text item_texts = list( itertools.chain.from_iterable( i.get_text().strip().split("\n") for i in items ) ) # ignore short lines item_texts = [ s for s in item_texts if len(s) >= 5 ] # generate the AslCard page_pos = 0 if items[0].y0 > lt_page.height/2 else 1 return AslCard( card_tag = _tidy( item_texts[0] ).replace( "# ", "#" ) , nationality = _tidy(item_texts[1]) if len(item_texts) > 1 else "" , name = _tidy(item_texts[2]) if len(item_texts) > 2 else "" , page_id = lt_page.pageid , page_pos = page_pos , ) def _extract_images( self , fname , max_pages ) : """Extract card images from a file.""" # extract each page from the PDF as an image fname_template = os.path.join( tempfile.gettempdir() , "asl_cards-%d.png" ) resolution = 300 # pixels/inch args = [ "_ignored_" , "-dQUIET" , "-dSAFER" , "-dNOPAUSE" , "-sDEVICE=png16m" , "-r"+str(resolution) , "-sOutputFile="+fname_template , ] if max_pages > 0 : args.append( "-dLastPage={}".format(max_pages) ) args.extend( [ "-f" , fname ] ) # FIXME! clean up left-over temp files before we start args = [ s.encode(locale.getpreferredencoding()) for s in args ] # FIXME! stop GhostScript from issuing warnings (stdout). ghostscript.Ghostscript( *args ) # figure out how many files were created (so we can show progress) npages = 0 for i in range(0,99999) : fname = fname_template % (1+i) if not os.path.isfile( fname ) : break npages += 1 # extract the cards from each page card_images = [] for page_no in range(0,npages) : if self.cancelling : raise RuntimeError("Cancelled.") # open the next page image self._progress2( float(page_no) / npages ) fname = fname_template % (1+page_no) img = Image.open( fname ) img_width , img_height = img.size # extract the cards (by splitting the page in half) fname2 = list( os.path.split( fname ) ) fname2[1] = os.path.splitext( fname2[1] ) ypos = img_height * 48 / 100 buf1 , size1 = self._crop_image( img , (0,0,img_width,ypos) , os.path.join( fname2[0] , fname2[1][0]+"a"+fname2[1][1] ) ) buf2 , size2 = self._crop_image( img , (0,ypos+1,img_width,img_height) , os.path.join( fname2[0] , fname2[1][0]+"b"+fname2[1][1] ) ) # check if this is the last page, and it has just 1 card on it if page_no == npages-1 and size1[1] < 1000 and size2[1] < 1000 : # yup - extract it buf , _ = self._crop_image( img , (0,0,img_width,img_height) , os.path.join( fname2[0] , fname2[1][0]+"a"+fname2[1][1] ) ) card_images.append( buf ) else : # nope - save the extracted cards card_images.append( buf1 ) card_images.append( buf2 ) # clean up os.unlink( fname ) return card_images def _crop_image( self , img , bbox , fname ) : # crop the image rgn = img.crop( bbox ) # trim the cropped region bgd_col = img.getpixel( (0,0) ) bgd_img = Image.new( img.mode , img.size , bgd_col ) diff = ImageChops.difference( rgn , bgd_img ) #diff = ImageChops.add(diff, diff, 2.0, -100) bbox = diff.getbbox() if bbox : rgn = rgn.crop( bbox ) # save the cropped image rgn.save( fname ) with open( fname , "rb" ) as fp : buf = fp.read() os.unlink( fname ) return buf , rgn.size def _progress( self , pval , msg ) : """Call the progress callback.""" if self.progress : self.progress( pval , msg ) def _progress2( self , pval ) : """Call the progress callback.""" if self.progress2 : self.progress2( pval ) # - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - _tidy_regex = re.compile( r"[,.()+-]" ) def _tidy( val ) : return _tidy_regex.sub(" ",val).strip()