1
0
Fork 0
flightgear/scripts/python/nasal_api_doc.py

296 lines
8.7 KiB
Python
Raw Normal View History

#!/usr/bin/env python
# -*- coding: utf-8 -*-
2012-03-04 18:49:28 +00:00
# Copyright (C) 2012 Adrian Musceac
#
# 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, see <http://www.gnu.org/licenses/>.
import os, sys, glob
import io
import re, string
"""Script which generates an API documentation file for Nasal libraries
located inside $FGROOT/Nasal/
2012-03-04 18:49:28 +00:00
Usage: nasal_api_doc.py parse [path to $FGROOT/Nasal/]
Or configure the local path below, and ommit the path in the console.
The API doc in HTML format is generated in the current working directory"""
########### Local $FGROOT/Nasal/ path ##########
NASAL_PATH="../fgfs/fgdata/Nasal/"
2012-03-04 18:49:28 +00:00
def get_files(nasal_dir):
if nasal_dir[-1]!='/':
nasal_dir+='/'
try:
os.stat(nasal_dir)
except:
print("The path does not exist")
sys.exit()
2012-02-03 17:16:11 +00:00
fgroot_dir = nasal_dir.rstrip('/').replace('Nasal','')
f_version = open(fgroot_dir+'version','rb')
version = f_version.read(256).rstrip('\n')
top_level = []
modules = []
top_namespaces = []
files_list = os.listdir(nasal_dir)
for f in files_list:
if f.find(".nas")!=-1:
top_level.append(f)
continue
if os.path.isdir(nasal_dir + f):
modules.append(f)
top_level.sort()
modules.sort()
if len(top_level) ==0:
print("This does not look like the correct $FGROOT/Nasal path")
sys.exit()
if len(modules)==0:
print("Warning: could not find any submodules")
for f in top_level:
namespace=f.replace(".nas","")
functions=parse_file(nasal_dir + f)
top_namespaces.append([namespace,functions])
for m in modules:
files=glob.glob(nasal_dir+m+"/*.nas")
for f in files:
functions=parse_file(f)
top_namespaces.append([m,functions])
2012-03-04 18:49:28 +00:00
2012-02-03 17:16:11 +00:00
output_text(top_namespaces,modules,version)
2012-03-04 18:49:28 +00:00
2012-02-03 17:16:11 +00:00
def output_text(top_namespaces,modules,version):
2012-03-04 18:49:28 +00:00
fw=open('./nasal_api_doc.html','wb')
buf='<html><head>\
<title>Nasal API</title>\
2012-02-03 17:16:11 +00:00
<style>\n\
a.main_module_link {margin-left:30px;display:block;float:left;}\
div.container {background-color:#eee;clear:left;margin-top:20px;}\
h2.namespace_title {padding-left:20px;color:#fff;background-color:#8888AC}\
h4.class_function {padding-left:20px;background-color:#eee;color:#000033}\
h4.class_definition {padding-left:20px;background-color:#eee;color:#000033}\
h4.function {padding-left:20px;background-color:#eee;color:#000033}\
hr {margin-left:30px;margin-right:30px;}\
div.comments {padding-left:40px;display:inline;font-size:12px;}\
</style>\n\
</head><body style="width:1024px;">'
2012-03-04 18:49:28 +00:00
2012-02-03 17:16:11 +00:00
buf+='<h1 style="padding-left:20px;display:block;color:#fff;background-color:#555588;">\
Nasal $FGROOT Library<br/><span style="font-size:12px;">Flightgear version: '+version+'\
2012-03-04 18:49:28 +00:00
<br/>This file is generated automatically by scripts/python/nasal_api_doc.py\
2012-02-03 17:16:11 +00:00
</span></h1>\
<br/><a href="http://plausible.org/nasal">Nasal documentation</a>&nbsp;&nbsp;\
<a href="http://wiki.flightgear.org/Nasal_scripting_language">Flightgear Nasal documentation</a>\n<div style="float:right;">&nbsp;'
buf+='<h2 style="font-size:14px;height:450px;width:250px;overflow:scroll;display:block;position:fixed;top:20px;right:20px;background-color:#8888AC;border:1px solid black;">\n'
done=[]
for namespace in top_namespaces:
color='0000cc'
if namespace[0] in modules:
color='cc0000'
if namespace[0] not in done:
2012-02-03 17:16:11 +00:00
buf+='<a class="main_module_link" style="color:'+color+'" href="#'+namespace[0]+'">'+namespace[0]+'</a>&nbsp;<br/>\n'
done.append(namespace[0])
buf+='</h2></div>\n'
done2=[]
for namespace in top_namespaces:
if namespace[0] not in done2:
2012-02-03 17:16:11 +00:00
buf+='<div class="container" style="">\n'
buf += '<h2 class="namespace_title"><a name="'+namespace[0]+'">'+namespace[0]+'</a></h2>\n'
done2.append(namespace[0])
for functions in namespace[1]:
class_func=functions[0].split('.')
if len(class_func)>1:
f_name=''
if class_func[1].find('_')==0:
f_name='<font color="#0000cc">'+class_func[1]+'</font>'
else:
f_name=class_func[1]
if class_func[1]!='':
2012-02-03 17:16:11 +00:00
buf+= '<div><h4 class="class_function"><b>'\
+namespace[0]+'</b>'+ "." + '<b><i>'+class_func[0]+'</i></b>'+'<b>.'+f_name+'</b>'+' ( <font color="#cc0000">'+ functions[1]+ '</font> )' +'</h4>\n'
else:
2012-02-03 17:16:11 +00:00
buf+= '<div><h4 class="class_definition"><b>'\
+namespace[0]+'</b>'+ "." + '<b><i><u><font color="#000000">'+class_func[0]+'</font></u></i></b>' +'</h4>\n'
else:
if functions[0].find('_')==0:
f_name='<font color="#0000cc">'+functions[0]+'</font>'
else:
f_name=functions[0]
2012-02-03 17:16:11 +00:00
buf+= '<div><h4 class="function"><b>'\
+namespace[0]+'</b>'+ "." + '<b>'+f_name+'</b>'+ ' ( <font color="#cc0000">'+ functions[1]+ '</font> )' +'</h4>\n'
for comment in functions[2]:
if comment.find('=====')!=-1:
2012-02-03 17:16:11 +00:00
buf+='<hr/>'
else:
2012-03-25 08:34:50 +00:00
tempComment = comment.replace('#','').replace('<','&lt;').replace('>','&gt;')
if tempComment.strip()!="":
2012-03-25 08:34:50 +00:00
buf+= '<div class="comments">'+tempComment+'</div><br/>\n'
buf+='</div>\n'
if namespace[0] not in done2:
2012-03-04 18:49:28 +00:00
buf+='</div>\n'
buf+='</body></html>'
fw.write(buf)
2012-03-04 18:49:28 +00:00
fw.close()
def parse_file(filename):
fr=open(filename,'rb')
content=fr.readlines()
i=0
retval=[]
classname=""
for line in content:
match=re.search(r'^var\s+([A-Za-z0-9_-]+)\s*=\s*func\s*\(?([A-Za-z0-9_\s,=.\n-]*)\)?',line)
if match!=None:
func_name=match.group(1)
comments=[]
param=match.group(2)
if(line.find(')')==-1 and line.find('(')!=-1):
k=i+1
while(content[k].find(')')==-1):
param+=content[k].rstrip('\n')
k+=1
param+=content[k].split(')')[0]
j=i-1
count=0
while ( j>i-35 and j>-1):
if count>3:
break
if len(content[j])<2:
j-=1
count+=1
continue
if re.search(r'^\s*#',content[j])!=None:
comments.append(content[j].rstrip('\n'))
j-=1
else:
break
if(len(comments)>1):
2012-03-04 18:49:28 +00:00
comments.reverse()
retval.append((func_name, param,comments))
i+=1
continue
2012-03-04 18:49:28 +00:00
match3=re.search(r'^var\s*([A-Za-z0-9_-]+)\s*=\s*{\s*(\n|})',line)
if match3!=None:
classname=match3.group(1)
2012-03-04 18:49:28 +00:00
comments=[]
2012-03-04 18:49:28 +00:00
j=i-1
count=0
while ( j>i-35 and j>-1):
if count>3:
break
if len(content[j])<2:
j-=1
count+=1
continue
if re.search(r'^\s*#',content[j])!=None:
comments.append(content[j].rstrip('\n'))
j-=1
else:
break
if(len(comments)>1):
2012-03-04 18:49:28 +00:00
comments.reverse()
retval.append((classname+'.', '',comments))
i+=1
continue
2012-03-04 18:49:28 +00:00
match2=re.search(r'^\s*([A-Za-z0-9_-]+)\s*:\s*func\s*\(?([A-Za-z0-9_\s,=.\n-]*)\)?',line)
if match2!=None:
func_name=match2.group(1)
comments=[]
param=match2.group(2)
if(line.find(')')==-1 and line.find('(')!=-1):
k=i+1
while(content[k].find(')')==-1):
param+=content[k].rstrip('\n')
2012-03-04 18:49:28 +00:00
k+=1
param+=content[k].split(')')[0]
j=i-1
count=0
while ( j>i-35 and j>-1):
if count>3:
break
if len(content[j])<2:
j-=1
count+=1
continue
if re.search(r'^\s*#',content[j])!=None:
comments.append(content[j].rstrip('\n'))
j-=1
else:
break
if(len(comments)>1):
2012-03-04 18:49:28 +00:00
comments.reverse()
if classname =='':
continue
retval.append((classname+'.'+func_name, param,comments))
i+=1
continue
2012-03-04 18:49:28 +00:00
match4=re.search(r'^([A-Za-z0-9_-]+)\.([A-Za-z0-9_-]+)\s*=\s*func\s*\(?([A-Za-z0-9_\s,=\n.-]*)\)?',line)
if match4!=None:
classname=match4.group(1)
func_name=match4.group(2)
comments=[]
param=match4.group(3)
if(line.find(')')==-1 and line.find('(')!=-1):
k=i+1
while(content[k].find(')')==-1):
param+=content[k].rstrip('\n')
k+=1
param+=content[k].split(')')[0]
j=i-1
count=0
while ( j>i-35 and j>-1):
if count>3:
break
if len(content[j])<2:
j-=1
count+=1
continue
if re.search(r'^\s*#',content[j])!=None:
comments.append(content[j].rstrip('\n'))
j-=1
else:
break
if(len(comments)>1):
2012-03-04 18:49:28 +00:00
comments.reverse()
retval.append((classname+'.'+func_name, param,comments))
i+=1
continue
2012-03-04 18:49:28 +00:00
i+=1
return retval
2012-03-04 18:49:28 +00:00
if __name__ == "__main__":
if len(sys.argv) <2:
print('Usage: nasal_api_doc.py parse [path to $FGROOT/Nasal/]')
sys.exit()
else:
if sys.argv[1]=='parse':
if len(sys.argv) <3:
nasal_path=NASAL_PATH
else:
nasal_path=sys.argv[2]
get_files(nasal_path)
else:
print('Usage: nasal_api_doc.py parse [path to $FGROOT/Nasal/]')
sys.exit()