ttbp/bin/core.py

438 lines
11 KiB
Python
Raw Normal View History

2016-04-30 01:56:40 +00:00
#!/usr/bin/python
2016-05-22 02:18:25 +00:00
'''
ttbp: tilde town blogging platform
(also known as the feels engine)
a console-based blogging program developed for tilde.town
copyright (c) 2016 ~endorphant (endorphant@tilde.town)
core.py:
this is a core handler for some ttbp standalone/output functions
Permission is hereby granted, free of charge, to any person obtaining
a copy of this software and associated documentation files (the
"Software"), to deal in the Software without restriction, including
without limitation the rights to use, copy, modify, merge, publish,
distribute, sublicense, and/or sell copies of the Software, and to
permit persons to whom the Software is furnished to do so, subject to
the following conditions:
The above copyright notice and this permission notice shall be
included in all copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE
LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
2016-05-22 02:18:25 +00:00
the complete codebase is available at:
https://github.com/modgethanc/ttbp
'''
2016-04-30 01:56:40 +00:00
import os
import time
2016-05-04 15:00:28 +00:00
import subprocess
import re
import mistune
import json
2016-05-01 22:26:26 +00:00
import chatter
2016-04-30 01:56:40 +00:00
2016-04-30 23:47:49 +00:00
SOURCE = os.path.join("/home", "endorphant", "projects", "ttbp", "bin")
USER = os.path.basename(os.path.expanduser("~"))
PATH = os.path.join("/home", USER, ".ttbp")
LIVE = "http://tilde.town/~"
WWW = os.path.join(PATH, "www")
2016-04-30 04:00:55 +00:00
CONFIG = os.path.join(PATH, "config")
2016-04-30 23:47:49 +00:00
DATA = os.path.join(PATH, "entries")
FEED = os.path.join(SOURCE, "www", "index.html")
DOCS = os.path.join(SOURCE, "www", "help.html")
NOPUB = os.path.join(CONFIG, "nopub")
SETTINGS = {}
2016-04-30 01:56:40 +00:00
2016-05-01 22:26:26 +00:00
HEADER = ""
FOOTER = ""
FILES = []
def load(ttbprc={}):
2016-05-22 02:18:25 +00:00
'''
get all them globals set up!!
'''
2016-05-01 22:26:26 +00:00
global HEADER
global FOOTER
global SETTINGS
2016-04-30 01:56:40 +00:00
2016-05-01 22:26:26 +00:00
HEADER = open(os.path.join(CONFIG, "header.txt")).read()
FOOTER = open(os.path.join(CONFIG, "footer.txt")).read()
SETTINGS = ttbprc
2016-04-30 01:56:40 +00:00
2016-05-01 23:34:20 +00:00
load_files()
def reload_ttbprc(ttbprc={}):
'''
reloads new ttbprc into current session
'''
global SETTINGS
SETTINGS = ttbprc
2016-05-01 23:34:20 +00:00
def load_files():
2016-05-22 02:18:25 +00:00
'''
file loader
* reads user's nopub file
* loads all valid filenames that are not excluded in nopub to global files list
'''
2016-05-01 23:34:20 +00:00
global FILES
FILES = []
2016-05-03 17:14:53 +00:00
for filename in os.listdir(DATA):
if nopub(filename):
link = os.path.join(WWW, os.path.splitext(os.path.basename(filename))[0]+".html")
if os.path.exists(link):
subprocess.call(["rm", link])
continue
2016-05-03 17:14:53 +00:00
filename = os.path.join(DATA, filename)
if os.path.isfile(filename) and valid(filename):
2016-05-03 17:14:53 +00:00
FILES.append(filename)
2016-04-30 01:56:40 +00:00
2016-05-01 22:26:26 +00:00
FILES.sort()
FILES.reverse()
2016-04-30 01:56:40 +00:00
## html outputting
2016-04-30 01:56:40 +00:00
def write(outurl="default.html"):
2016-05-22 02:18:25 +00:00
'''
main page renderer
* takes everything currently in FILES and writes a single non-paginated html
file
* calls write_page() on each file to make permalinks
'''
2016-04-30 01:56:40 +00:00
outfile = open(os.path.join(WWW, outurl), "w")
outfile.write("<!--generated by the tilde.town blogging platform on "+time.strftime("%d %B %y")+"\nhttp://tilde.town/~endorphant/ttbp/-->\n\n")
2016-04-30 01:56:40 +00:00
for line in HEADER:
outfile.write(line)
outfile.write("\n")
2016-05-03 17:14:53 +00:00
for filename in FILES:
write_page(filename)
for line in write_entry(filename):
2016-04-30 01:56:40 +00:00
outfile.write(line)
outfile.write("\n")
2016-04-30 01:56:40 +00:00
for line in FOOTER:
outfile.write(line)
outfile.close()
2016-04-30 23:47:49 +00:00
return os.path.join(LIVE+USER,os.path.basename(os.path.realpath(WWW)),outurl)
2016-04-30 04:00:55 +00:00
2016-05-03 17:14:53 +00:00
def write_page(filename):
2016-05-22 02:18:25 +00:00
'''
permalink generator
* makes a page out of a single entry for permalinking, using filename/date as
url
'''
2016-05-03 17:14:53 +00:00
outurl = os.path.join(WWW, "".join(parse_date(filename))+".html")
outfile = open(outurl, "w")
outfile.write("<!--generated by the tilde.town blogging platform on "+time.strftime("%d %B %y")+"\nhttp://tilde.town/~endorphant/ttbp/-->\n\n")
for line in HEADER:
outfile.write(line)
outfile.write("\n")
for line in write_entry(filename):
outfile.write(line)
outfile.write("\n")
for line in FOOTER:
outfile.write(line)
outfile.close()
return outurl
def write_entry(filename):
2016-05-22 02:18:25 +00:00
'''
entry text generator
* dump given file into entry format by parsing file as markdown
* return as list of strings
'''
2016-04-30 01:56:40 +00:00
2016-05-03 17:14:53 +00:00
date = parse_date(filename)
2016-04-30 02:09:30 +00:00
2016-04-30 01:56:40 +00:00
entry = [
2016-04-30 02:09:30 +00:00
"\t\t<p><a name=\""+date[0]+date[1]+date[2]+"\"></a><br /><br /></p>\n",
2016-04-30 01:56:40 +00:00
"\t\t<div class=\"entry\">\n",
"\t\t\t<h5><a href=\"#"+date[0]+date[1]+date[2]+"\">"+date[2]+"</a> "+chatter.month(date[1])+" "+date[0]+"</h5>\n"
#"\t\t\t<P>"
2016-04-30 01:56:40 +00:00
]
raw = []
2016-05-03 17:14:53 +00:00
rawfile = open(os.path.join(DATA, filename), "r")
2016-04-30 01:56:40 +00:00
for line in rawfile:
raw.append(line)
rawfile.close()
entry.append("\t\t\t"+mistune.markdown("".join(raw), escape=False, hard_wrap=False))
2016-04-30 01:56:40 +00:00
#for line in raw:
#entry.append(line+"\t\t\t")
#if line == "\n":
# entry.append("</p>\n\t\t\t<p>")
#entry.append("</p>\n")
2016-05-03 17:14:53 +00:00
entry.append("\t\t\t<p style=\"font-size:.6em; font-color:#808080; text-align: right;\"><a href=\""+"".join(date)+".html\">permalink</a></p>\n")
entry.append("\n\t\t</div>\n")
2016-04-30 01:56:40 +00:00
return entry
def write_global_feed(blogList):
2016-05-22 02:18:25 +00:00
'''
main ttbp index printer
2016-05-22 02:18:25 +00:00
* sources README.md for documentation
* takes incoming list of formatted blog links for all publishing blogs and
prints to blog feed
2016-05-22 02:18:25 +00:00
'''
2016-04-30 01:56:40 +00:00
outfile = open(FEED, "w")
2016-04-30 02:09:30 +00:00
## header
outfile.write("""\
<!DOCTYPE html PUBLIC \"-//W3C//DTD HTML 3.2//EN\">
<html>
<head>
<title>tilde.town feels engine</title>
<link rel=\"stylesheet\" href=\"style.css\" />
</head>
<body>
<div class="meta">
<h1>tilde.town feels engine</h1>
2016-04-30 02:09:30 +00:00
<h2><a href="https://github.com/modgethanc/ttbp">github
repo</a> | <a
href="http://tilde.town/~endorphant/blog/20160510.html">state
of the ttbp</a></h2>
<!--<p>curious? run <b>~endorphant/bin/ttbp</b> while logged in to tilde.town.</p>
<p>it's still a little volatile. let me know if anything breaks.</p>---></div>
<p>&nbsp;</p>
""")
## docs
outfile.write("""\
<div class="docs">""")
outfile.write(mistune.markdown(open(os.path.join(SOURCE, "..", "README.md"), "r").read()))
outfile.write("""\
</div>""")
## feed
outfile.write("""\
<p>&nbsp;</p>
<div class=\"feed\">
<h3>live feels-sharing:</h3>
<ul>""")
for blog in blogList:
outfile.write("""
<li>"""+blog+"""</li>\
""")
## footer
outfile.write("""
</ul>
</div>
</body>
</html>
""")
outfile.close()
## misc helpers
2016-05-04 15:00:28 +00:00
def meta(entries = FILES):
2016-05-22 02:18:25 +00:00
'''
metadata generator
* takes a list of filenames and returns a 2d list:
[0] absolute path
[1] mtime
2016-05-22 02:18:25 +00:00
[2] wc -w
[3] timestamp "DD month YYYY at HH:MM"
[4] entry date YYYY-MM-DD
[5] author
* sorted in reverse date order by [4]
'''
2016-05-04 15:00:28 +00:00
meta = []
for filename in entries:
mtime = os.path.getmtime(filename)
try:
wc = subprocess.check_output(["wc","-w",filename], stderr=subprocess.STDOUT).split()[0]
except subprocess.CalledProcessError:
wc = "???"
timestamp = time.strftime("%Y-%m-%d at %H:%M", time.localtime(mtime))
2016-05-04 15:00:28 +00:00
date = "-".join(parse_date(filename))
author = os.path.split(os.path.split(os.path.split(os.path.split(filename)[0])[0])[0])[1]
meta.append([filename, mtime, wc, timestamp, date, author])
2016-05-04 15:00:28 +00:00
meta.sort(key = lambda filename:filename[4])
meta.reverse()
2016-05-04 15:00:28 +00:00
return meta
def valid(filename):
2016-05-22 02:18:25 +00:00
'''
filename validator
* check if the filename is YYYYMMDD.txt
'''
filesplit = os.path.splitext(os.path.basename(filename))
if filesplit[1] != ".txt":
return False
pattern = '^((19|20)\d{2})(0[1-9]|1[0-2])(0[1-9]|1\d|2\d|3[01])$'
if not re.match(pattern, filesplit[0]):
return False
return True
def parse_date(file):
2016-05-22 02:18:25 +00:00
'''
parses date out of pre-validated filename
2016-05-22 02:18:25 +00:00
* assumes a filename of YYYYMMDD.txt
* returns a list:
[0] 'YYYY'
[1] 'MM'
[2] 'DD'
2016-05-22 02:18:25 +00:00
'''
rawdate = os.path.splitext(os.path.basename(file))[0]
date = [rawdate[0:4], rawdate[4:6], rawdate[6:]]
return date
def find_ttbps():
'''
returns a list of users with a ttbp by checking for a valid ttbprc
'''
users = []
for townie in os.listdir("/home"):
if os.path.exists(os.path.join("/home", townie, ".ttbp", "config", "ttbprc")):
users.append(townie)
return users
def publishing(username = USER):
'''
checks .ttbprc for whether or not user opted for www publishing
'''
ttbprc = {}
if username == USER:
ttbprc = SETTINGS
else:
ttbprc = json.load(open(os.path.join("/home", username, ".ttbp", "config", "ttbprc")))
return ttbprc.get("publishing")
def www_neighbors():
'''
takes a list of users with publiishing turned on and prepares it for www output
'''
userList = []
for user in find_ttbps():
if not publishing(user):
continue
userRC = json.load(open(os.path.join("/home", user, ".ttbp", "config", "ttbprc")))
url = ""
if userRC["publish dir"]:
url = LIVE+user+"/"+userRC["publish dir"]
lastfile = ""
files = os.listdir(os.path.join("/home", user, ".ttbp", "entries"))
files.sort()
for filename in files:
if valid(filename):
lastfile = os.path.join("/home", user, ".ttbp", "entries", filename)
if lastfile:
last = os.path.getctime(lastfile)
timestamp = time.strftime("%Y-%m-%d at %H:%M", time.localtime(last)) + " (utc"+time.strftime("%z")[0]+time.strftime("%z")[2]+")"
else:
timestamp = ""
last = 0
userList.append(["<a href=\""+url+"\">~"+user+"</a> "+timestamp, last])
# sort user by most recent entry
userList.sort(key = lambda userdata:userdata[1])
userList.reverse()
sortedUsers = []
for user in userList:
sortedUsers.append(user[0])
write_global_feed(sortedUsers)
def nopub(filename):
'''
checks to see if given filename is in user's NOPUB
'''
exclude = []
if os.path.isfile(NOPUB):
for line in open(NOPUB, "r"):
exclude.append(line.rstrip())
return os.path.basename(filename) in exclude
#############
#############
#############
2016-05-04 15:00:28 +00:00
def test():
load()
metaTest = meta()
for x in metaTest:
print(x)