The goal of this code is to print the entity tree of a VHDL project. There are a readme and very minimal tests on the github repo.

I am trying to refactor the code to use generators in order to familiarize myself with yield, and to think in terms of iterators, not raw data structures. Here is my current attempt at doing so:

import re
from sys import argv
from os import walk
from os.path import join as pjoin

EXCLUDES = ["implementation", "testbench"]
BASIC_ID_REGEX = "[a-z][a-z0-9]*(?:_[a-z0-9]+)*"

def _vhdltree(level, filepath, pattern, vhd_files):
    for entity, component in find_entities(filepath, pattern):
        """Codereview: I am not specifically interested in feedbacks for the
        following snippet (except about the recursive design), but if you 
        have a particularly elegant
        solution which keeps the "UI" part very minimal, suggestions
        are welcome"""
        path = vhd_files.get(component.lower(), "Not Found")
        print("    "*level + entity + " : " + path)
        if path != "Not Found":#Probably ugly, but lazy
            _vhdltree(level+1, path, pattern, vhd_files)

def find_entities(filepath, pattern):
    with open(filepath) as f:
        for l in f:
            m = pattern.match(l)
            if m:
                yield m.group('entity'), m.group('component').split(".")[-1]

def find_vhd(proot):
    for (dirpath, dirnames, filenames) in walk(proot):
        if not isexcluded(dirpath.lower()):
            for fn in filenames:
                if fn[-4:].lower() == ".vhd":
                    yield fn[:-4].lower(), pjoin(dirpath, fn)

def isexcluded(path):
    for excluder in EXCLUDES:
        if excluder in path:
            return True
    return False

def vhdltree(filepath, proot):
    instantiation_regex = ("\s*(?P<entity>{0})\s*:\s*entity\s*(?P<component>{0}(?:\.{0})*)"  # NOQA
    p = re.compile(instantiation_regex, re.IGNORECASE)
    vhd_files = dict(find_vhd(proot))
    _vhdltree(0, filepath, p, vhd_files)

if __name__ == "__main__":
    #Codereview: user is trusted, don't care about this part
    vhdltree(argv[1], argv[2])

Any feedback about the use of generators is welcome. Specifically, I wonder if it would be possible to avoid "realizing" the vhd_files generator as a dict while keeping the search in _vhdltree as simple as a get? Are there more pythonic alternatives or more efficient solutions for a structure designed to be one-wa searchable (I know this is probably never where the program spends its time, but I'd like to know for my general culture). I looked in itertools but didn't find an immediate solution.

I am interested in modern, Python3 solutions for my general knowledge. From a practical point of view however, the code is supposed to be 2-compatible.

  • 1
    \$\begingroup\$ Welcome to Code Review. I hope you get good answers. \$\endgroup\$
    – Tolani
    Oct 13, 2016 at 12:19
  • \$\begingroup\$ Please have a look at the help-center to see what you are and what you are not allowed to do, once your question has received an answer. You can ask a new question to ask about your updated code. \$\endgroup\$
    – Graipher
    Oct 24, 2016 at 10:38

1 Answer 1



Since you were asking about generators specifically, let’s start with some suggestions on how you’re using them:

  • Your find_entities() function opens a file, then iterates over the lines in that file. That’s a good use of Python’s ability to iterate directly over the lines in a file, but it’s a bit fiddly to unit test because you have to create files.

    I often write my generators to take an iterable of lines (which could be an open file object, or something else) – this makes the code easier to unit test, and more generic. I can read the lines from an arbitrary iterator, not just a file. It’s a small change:

    def find_entities(lines, pattern):
        for line in lines:
            m = pattern.match(line)
            if m:
                yield m.group('entity'), m.group('component').split('.')[-1]

  • Your _vhdltree() function is really doing two things: it’s working out the level, entity and path for an element in the iterator, and then printing it. Problem with printing is that it becomes hard to reuse this computation elsewhere. Better if you split the printing out from this function, and defer that to the caller.

    Something like:

    def parse_vhdltree(level, filepath, pattern, vhd_files):
        for entity, component in find_entities(open(file path), vhd_files):
                path = vhd_files[component.lower()]
            except KeyError:
            yield level, entity, path
            for entry in parse_vhdltree(level+1, path, pattern, vhd_files):
                yield entry
    for level, entity, path in parse_vhdltree(0, filepath, p, vhd_files):
        print('    ' * level + entity + ' : ' + path)

    (I’ll talk about the KeyError in a minute.)

  • I really like your find_vhd() generator, and it’s a good example of how generators can clean up code in Python. All the messy logic of deciding which files to count is absorbed in this function.

    One tweak I would make is to throw away the dirnames data – you’re not using it. Using an underscore in its place is a common convention to indicate you’re not interested in a particular value:

    def find_vhd(proot):
        for (dirpath, _, filenames) in walk(proot):
            if not isexcluded(dirpath.lower()):
                for fn in filenames:
                    if fn[-4:].lower() == ".vhd":
                        yield fn[:-4].lower(), pjoin(dirpath, fn)

General Python style

Generators aside, here are some comments on your general Python style:

  • I have a bit of a hobby horse about use of regular expressions. I think you’re using them for a sensible reason, but you could be better about commenting, and explaining what they’re trying to match.

    Breaking it over multiple lines, and explaining exactly what you’re trying to match against, will make it much easier to read. Additionally, when you come back to this code in six months, you’ll thank your past self for those comments.

    I would also suggest pre-compiling the instantiation_regex at the top of the file, alongside BASIC_ID_REGEX.

  • Your isexcluded() function could be replaced with a one-liner:

    return any(excluder in path for excluder in EXCLUDES)

    This is using a lazy generator expression as the argument, and the nice thing about any() is that it exits as soon as it gets a positive element.

    I see that you’re already using an equivalent expression with all() in the GitHub version, which is similarly lazy. It exits as soon as it finds a negative element.

  • I’m not a super fan of from foo import bar style imports, especially from the standard library. Even worse if from foo import bar as baz, as you’ve done with os.path.join.

    Characters are cheap – use the full, stdlib name. It will make your code much easier to read for other Python developers, who don’t need to worry about what pjoin() does. They can just recognise os.path.join.

  • Rather than using a sentinel value from dict.get() to denote a missing key in the dictionary, do a lookup with square brackets and catch the KeyError. This is more Pythonic, and avoids problems if the sentinel value turns up in the dictionary.

    And I think using a dictionary here is fine. It’s simple, clean, and it works.

  • \$\begingroup\$ Thank you very much for your feedback. I'll probably end up applying most of it (or would if this wasn't a toy project). As my update shows, I diverged a bit from your suggestions. I really like your use of yield in the main recursive part, but I don't really like the fact that you explicitly 3 things at one point and then a tuple as a single thing later on (It must not be very readable, because I myself took 10seconds to figure out what as going on). \$\endgroup\$
    – nathdwek
    Oct 15, 2016 at 9:44
  • \$\begingroup\$ Modern Python3 has yield from which is really really elegant imo (see edit), but in production code I will still apply this while explicitly yielding three things (see github). \$\endgroup\$
    – nathdwek
    Oct 15, 2016 at 9:45
  • \$\begingroup\$ I used the try except method to handle paths and I like it more, but I used else rather than continue to still components that were not found (actually important information). Moreover, I have been unfortunately formatted to not like continue, break et al, but I still objectively think that else is a tiny bit more elegant here. \$\endgroup\$
    – nathdwek
    Oct 15, 2016 at 9:47
  • \$\begingroup\$ Again, many thanks for the feedback. Do you have 5 mins to review the updated version in my post? \$\endgroup\$
    – nathdwek
    Oct 15, 2016 at 9:48

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.