Comments by "Lawrence D’Oliveiro" (@lawrencedoliveiro9104) on "Tech With Tim" channel.

  1. About talking you are what.
    26
  2. 12:58 There is a chance that could fail. asyncio only holds weak references to the tasks it creates; it relies on the caller (you) to maintain strong references, to stop those tasks from disappearing randomly. Here it executes after the “task” variable holding the reference has gone.
    5
  3. 8:13 Have you tried using a Jupyter notebook? It’s good for “scratchpad”-style programming. Which means I can leave the driver object open and incrementally run extra commands against it, instead of having to close and reopen it every time.
    5
  4. Threading and multiprocessing are preemptive, coroutines are not. Coroutines and threading are default-share-everything, multiprocessing is default-share-nothing. Coroutines are easier to debug than threads.
    4
  5. 7:35 Worth mentioning at this point the difference between “stackless” and “stackful” coroutines. Python has the “stackless” kind, which is why the await keyword can only be used in the top level of an async-function, not anywhere else.
    4
  6. 1:48 Debian includes standard packages for python3-selenium, chromium-driver and phantomjs. Both the latter two are listed as dependencies for the first one.
    4
  7. C has no “smart pointers”, but you can manage memory more reliably using structured programming and avoiding gotos. https://github.com/ldo/a_structured_discipline_of_programming
    2
  8. This is not really parallelism. There is only one thread of execution. It simply gives you a way to respond to outside events nondeterministically.
    2
  9. The logging module just gives you more control over message output, that’s all. If you use sys.stderr.write, then you have to do your own implementation of verbosity levels, for example. The standard Python module has already pre-invented this whole wheel (and much else besides), so you don’t have to.
    1
  10.  @SimGunther  That’s not about cache locality, it’s about correctness.
    1
  11. Extreme UML Waterfall will be the new hawtness. Get in on the ground floor now.
    1
  12. 11:08 I would take out the try-clause, and just leave the WebDriverWait call. If it’s going to raise an exception, it can be instructive for tutorial purposes to see what that might be. Leave the error recovery for a more “production-quality” script.
    1
  13.  @alexanderelgert6037  Is it a ratio or does it have units? Just wondering if all code complexity is necessarily cyclomatic ...
    1
  14. I can program Emacs to jump between lines with matching indentation. This makes it very easy to navigate around Python code, provided you have end markers for compound statements, e.g.: if ... ... #end if while ... ... #end while These provide the targets to jump to, to take you to the end of each construct. https://ldo17.tumblr.com/post/155087402702/fixing-pythons-greatest-mistake
    1
  15. Java’s not that fast. Things slow down spectacularly once the garbage collector kicks in. Also it’s a memory hog. Python (or at least CPython) reduces that by doing reference-counting so most objects disappear the moment the last reference to them is forgotten, with full garbage collection only needed in more difficult cases.
    1
  16. How about a recursive generator? Here’s one that yields all permutations of a given sequence: def permute(seq) : "generator which yields successive permutations of the elements of seq." if len(seq) == 0 : yield () else : for i in range(0, len(seq)) : for rest in permute(seq[:i] + seq[i + 1:]) : yield (seq[i],) + rest ♯end for ♯end for ♯end if ♯end permute Example use: >>> list(permute(("pears", "oranges", "feijoas"))) [('pears', 'oranges', 'feijoas'), ('pears', 'feijoas', 'oranges'), ('oranges', 'pears', 'feijoas'), ('oranges', 'feijoas', 'pears'), ('feijoas', 'pears', 'oranges'), ('feijoas', 'oranges', 'pears')]
    1
  17. 3:27 Notice we have both “find_elements_by_class_name” and “find_element_by_class_name”. I suspect what you said earlier about only finding the first instance of an element with a specified class applies to the latter call, not to the former; the former one would let you iterate through them all.
    1
  18. 3:49 I just tried this: list(e.get_attribute("name") for e in driver.find_elements_by_class_name("search-field")) and got back ['s']
    1
  19. I know that feeling. Particularly fun with a sleep call: await asyncio.sleep(«n») vs asyncio.sleep(«n»)
    1
  20. Did you know that asyncio is implemented in pure Python? That the whole task-dispatching system works through an interesting (deliberately-designed) interaction between coroutines and generators? I wrote up a description of this in a Jupyter notebook, which I cannot link to here.
    1
  21. The second version is fine, too. Note that the docs state that no formatting substitution is done on the message string if no formatting arguments are supplied. Compare, say, printf(3), which always does formatting substitution even when it doesn’t make sense. Python is a bit smarter.
    1
  22. Functions that are useful in multiple places are a sign of effective code reuse.
    1
  23.  @gJonii  Look at, for example, the built-in integer operations. Do they have a “clear domain where they operate in”? Or are they in fact widely used in a lot of different code?
    1
  24.  @gJonii  So you feel Python shouldn’t have such operations, because you find them “confusing”?
    1
  25.  @gJonii  You need to stop trying to describe something as fundamental to developer productivity as code reuse as an “antipattern”.
    1
  26. What about the ones in the standard library? Lots of good stuff there. For example, ctypes -- anybody else made good use of that? https://github.com/topics/ctypes-wrapper
    1