astlan-diff.py
- This topic has 6 replies, 3 voices, and was last updated 4 years, 6 months ago by
.
-
Posts
-
I have been adding names and some places to the dictionary, that way if I misspell a name it will be in red.
I have also put the more complicated ones in autotext so I don’t have to type them all.
I am not happy with any of the diff options I am playing with. If there was a good free tool that people could compare PDF or epubs with that was easy for non-tech people to use that would be good, but not seeing it, and not sure how much time its worth.
What I might just do is publish a PDF version with revisions/markup showing. I’ve actually had to turn this off to get “clean” alpha PDF’s.
Once that is published. I accept all the changes, which starts the clock anew (print the clean version to pdf/epub) and then start tracking again for the next release.
A Python script for finding changes between Astlan EPUB editions:
[code=python]
#!/usr/bin/python3
# -*- coding: utf-8 -*-import difflib
import epub
from bs4 import BeautifulSoupdef text_paras(book, href):
soup = BeautifulSoup(book.read_item(href), ‘lxml’)
return [x.get_text() for x in soup.find_all(‘p’)]book1 = epub.open_epub(‘Apostles of Doom Alpha 1 – J. L. Langland.epub’)
book2 = epub.open_epub(‘Apostles of Doom Alpha 2 – J. L. Langland.epub’)for item in book1.opf.manifest.values():
print(item.href)
if item and item.media_type == ‘application/xhtml+xml’:
paras1 = text_paras(book1, item.href)
paras2 = text_paras(book2, item.href)s = difflib.SequenceMatcher(None, paras1, paras2)
for opcode in s.get_opcodes():
print(“%6s a[%d:%d] b[%d:%d]” % opcode)if opcode[0] == ‘insert’:
print(‘B: ‘, paras2[opcode[3]:opcode[4]])elif opcode[0] == ‘replace’:
print(‘A: ‘, paras1[opcode[1]:opcode[2]])
print(‘B: ‘, paras2[opcode[3]:opcode[4]])elif opcode[0] == ‘delete’:
print(‘A: ‘, paras1[opcode[1]:opcode[2]])[/code]
Very interesting!
Great minds think a like, before I read this I just posted a reply and asked people if a diff document would be useful, something showing changes, etc.
I was also looking at using either NLTK or Google’s Tensor Flow language models to pull out all of the character and place names for spell checking, but that will have to wait until later.
- You must be logged in to reply to this topic.