I’m doing some writing on the topic of hybrid automata, and it looks like the best way to generate figures of the automata in LaTeX is with GraphViz and the dot2texi package, which works with the very nice dot2tex tool to generate figures from GraphViz dot formatted code embedded directly in LaTeX.

One of the issues, however, is that it expects an executable called “dot2tex” to be available on the shell. This means that, first and foremost, shell escape needs to be enabled when compiling the document (–shell-escape in many distributions, and –enable-write18 in MiKTeX, which I use). Adding this line to the beginning of the LaTeX source will handle it:

%& --shell-escape --enable-write18

Next, since the executable file that’s installed (in Windows on Python 2.7 it goes, by default, to C:\Python27\Scripts\dot2tex) is made executable by a shebang on its first line, even if this file is on your path in Windows, it won’t do anything. So you need to use ExeMaker, which creates a very lightweight exe file to call a Python script. I put it in a directory that’s in my PATH (I made C:\Users\george\bin for this purpose) ran it like this:

C:\Users\george> rename C:\Python27\Scripts\dot2tex dot2tex.py
C:\Users\george> exemaker C:\Python27\Scripts\dot2tex.py

In my case, this created dot2tex.exe and dot2tex.py in C:\Users\george\bin. However, running dot2tex.exe causes an error:

C:\Users\george>dot2tex
Traceback (most recent call last):
  File "C:\Users\george\bin\dot2tex.py", line 6, in <module>
    from dot2tex.dot2tex import main
  File "C:\Users\george\bin\dot2tex.py", line 6, in <module>
    from dot2tex.dot2tex import main
ImportError: No module named dot2tex

A bit of poking around reveals that this is due to the way that Python searches paths in import statements: the script, dot2tex.py, conflicts in its name with the module that we want to import, and the first directory that Python searches when looking for a file to import is the directory containing the executing script (or, in interactive mode, the working directory). The workaround I came up with, which seems to work well, is just to add two lines to dot2tex.py:

import sys
del sys.path[0]

Now my dot2tex.py file (the one in C:\Users\george\bin) looks like this:

#!C:\Python27\python.exe
import site
import sys
del sys.path[0]
#!C:\Python27\python.exe
from dot2tex.dot2tex import main
if __name__ == '__main__':
    main()

So far, this seems to have fixed all of my problems.

I love the site but wanted to branch out a little bit into some samples outside of the PHP and WordPress world. In my first attempt, I didn’t branch very far, and it’s a bit longer than I would have preferred, but here it is — from MoinMoin. Hopefully I included enough information to find the issue.

def getUserId(request, searchName):
    """
    Get the user ID for a specific user NAME.
 
    @param searchName: the user name to look up
    @rtype: string
    @return: the corresponding user ID or None
    """
    if not searchName:
        return None
    cfg = request.cfg
    try:
        _name2id = cfg._name2id
    except AttributeError:
        arena = 'user'
        key = 'name2id'
        cache = caching.CacheEntry(request, arena, key)
        try:
            _name2id = pickle.loads(cache.content())
        except (pickle.UnpicklingError, IOError, EOFError, ValueError):
            _name2id = {}
        cfg._name2id = _name2id
    id = _name2id.get(searchName, None)
    if id is None:
        for userid in getUserList(request):
            name = User(request, id=userid).name
            _name2id[name] = userid
        cfg._name2id = _name2id
        arena = 'user'
        key = 'name2id'
        cache = caching.CacheEntry(request, arena, key)
        cache.update(pickle.dumps(_name2id, PICKLE_PROTOCOL))
        id = _name2id.get(searchName, None)
    return id
 
class User:
    """A MoinMoin User"""
 
    def __init__(self, request, id=None, name="", password=None,
                 auth_username="", **kw):
        """ Initialize User object
 
        @param request: the request object
        @param id: (optional) user ID
        @param name: (optional) user name
        @param password: (optional) user password (unicode)
        @param auth_username: (optional) already authenticated user name
                              (e.g. when using http basic auth) (unicode)"""
        self._cfg = request.cfg
        self.valid = 0
        self.trusted = 0
        self.id = id
        self.auth_username = auth_username
        self.auth_method = kw.get('auth_method', 'internal')
        self.auth_attribs = kw.get('auth_attribs', ())
 
        # we got an already authenticated username:
        check_pass = 0
        if not self.id and self.auth_username:
            self.id = getUserId(request, self.auth_username)
            if not password is None:
                check_pass = 1
        if self.id:
            self.load_from_id(check_pass)
            if self.name == self.auth_username:
                self.trusted = 1
        elif self.name:
            self.id = getUserId(self._request, self.name)
            if self.id:
                self.load_from_id(1)
            else:
                self.id = self.make_id()
        else:
            self.id = self.make_id()
 
    def __filename(self):
        """ Get filename of the user's file on disk
        @rtype: string
        @return: full path and filename of user account file
        """
        return os.path.join(self._cfg.user_dir, self.id or "...NONE...")
 
    def save(self):
        """ Save user account data to user account file on disk.
 
        This saves all member variables, except "id" and "valid" and
        those starting with an underscore.
        """
        if not self.id:
            return
 
        user_dir = self._cfg.user_dir
        filesys.makeDirs(user_dir)
 
        self.last_saved = str(time.time())
 
        data = codecs.open(self.__filename(), "w", config.charset)
        data.write("# Data saved '%s' for id '%s'\n" % (
            time.strftime(self._cfg.datetime_fmt, time.localtime(time.time())),
            self.id))
        attrs = vars(self).items()
        attrs.sort()
        for key, value in attrs:
            if key not in self._cfg.user_transient_fields and key[0] != '_':
                # Encode list values
                if key in ['quicklinks', 'subscribed_pages']:
                    value = encodeList(value)
                line = u"%s=%s\n" % (key, unicode(value))
                data.write(line)
        data.close()
 
        if not self.disabled:
            self.valid = 1

Java is not Python. Java’s changes that introduced generics (to allow expressions like, for example, “Vector<Integer>”) were pretty much all in the compiler. The bytecode stays the same; the compiler just gets some additional data to use for typechecking. Then your beautifully generic Java 5.0 code undergoes a process called type erasure producing the same — but better typechecked — bytecode that your old 1.4 code would have. The result is that when I, with my limited understanding of Java generics, tried to write overly dynamic code, the compiler scoffed.

“required: class or array”, it said. What it really meant was, “You can’t use a generic type in an instanceof operation, you dummy, because I’m about to erase that type and replace it with Object.” Doi.

Cygwin is not Linux. Especially where Java is concerned. I spent about 20 minutes trying to figure out why I couldn’t pass the JDK an absolute classpath from Cygwin when relative classpaths worked just fine. It was because the JDK I was calling is Windows native, not Cygwin, so it doesn’t understand the Unix-style directory structure I was giving it.

Java is not Python (redux). I’ve mostly been using Python for my various and sundry programming needs over the past couple of years. And it’s impressive how quickly one gets used to a REPL (read-eval-print-loop) — and how useful it is for quick debugging and answers.

For instance, I missed having a REPL when I had a simple question: can I typecast null to whatever object type I want (the answer is yes, by the way, which makes a lot of sense when you think about it)? I was in the midst of a fairly significant rewrite so I couldn’t just try it and run it. So I saw four choices: assume it would work and test it later, assume it wouldn’t work and rewrite it, look it up, or write a Java program to test it.

Turns out that there’s actually a fifth option: use the Java REPL. It’s called BeanShell (or bsh), and I feel pretty much ridiculous for not having found it before. Maybe you’ll enjoy it.

Okay, back to work.