Comments by "Dr Gamma D" (@DrDeuteron) on "Tech With Tim" channel.

  1. the dataclass module also has astuple() and asdict() functions that are really useful when you're sending your data to another function (e.g., a plotter) that doesn't care about all the class structure. And the "field" function is great for defining what each attribute is. I use it on space craft telemetry, and you can attach ranges, alarm levels, units (ounce-inch or N-m, for example), and also LaTex r-strings for status plots. It saves soooooo much boiler plate code. Moreover, it puts all the information in static variables attached to the instance, not a module dictionary, or worse: run-time magic numbers .
    13
  2. I was given a csv file with 524 columns of date times, instrument telemetry, etc...dataclass saved the day...just cut and paste the column header names and let the decorator do the rest. Also put the in-spec limits in the field's metadata along with LaTex strings for y-axis labels. 524...times. data class ftw.
    10
  3. python lore states "if you think you need metaclasses, you don't".
    7
  4. Nested list comp: >>>flattened = list(chain(*matrix)) chain is from itertools.
    7
  5. for me, when I'm reading 545 data fields of space craft telemetry, which is a helluva dunder init. Instead I just cut and paste the attributes from an excel file into a data class and it calls for a single mouse click. Moreover, dataclass comes with "fields", which can be used to describe the fields, the LaTex formulas for them. expected ranges, alarm ranges, broke ranges, and there is an astuple functions, which can be stuck in dunder iter, possible zip with the fields, and in about 3 lines of executable code, you can flag out-of-range telemetry. Very little dynamic code, and lots of static could is much simpler and less bug prone.
    3
  6. "is not False" "!= False" is unpythonic, But since the default is None: val.get("someKey") is not None is preferred, unless None is an expected value ofc.
    3
  7.  @TohaBgood2  I've use multiprocessing, and it worked. But I don't know how or why.
    3
  8. Overthinking: lambda x: x**2 is "f(x) = x**2" wherever it appears.
    3
  9. regarding dunder iter, everything you said was correct, but may be unclear to noobs. You don't need a next method if you're not tracking internal state, so you can out the loop directly in dunder iter: def __iter__(self): for value in range(self.current, self.end): yield value works fine, and iter(...) returns a generator which manages the StopIteration for you. Now if you need to monitor state, say you're iterating over a list files, then use dunder next and other methods can see self.current in-between next() calls.
    2
  10. I don't think enumerate returns an index, as evidenced by the "start=" keyword..rather it provides a count. Useful for managing matplotlib subplots, for example.
    2
  11. You called dunder str / repr "descriptors", when they are really 'descriptions'. Descriptors involve dunder get / set, and a video on when/why to use those would be fantastic.
    2
  12.  @rpitit  shoulda just replied AttributeError: 'dict' object has no attribute 'age'
    2
  13. is any()'s evaluation short-circuited?
    2
  14.  @sarahalzaher7474  no, clever.
    2
  15.  @JakeCallahan in 2, new style classes had to sub object. Without that parameter, they were old style classes, which were sunset before I started python in 2002, but still existed in the language for backwards compatibility.
    2
  16. idk, I make satellites, and we got QA all up in our business,
    2
  17. on post init: beware doing work in the constructor....
    1
  18. the mistake here is making and instance attribute, 'val', that can't be instantiated. Default to a class attribute and automatically create and instance attribute when accesed. (though f is a sin for modifying an instance variable outside an "h" method) >>> class h: ... val = 0 ... >>> def f(g=h()): ... g.val+=1 ... return g.val ... >>> f() 1 >>> f() 2 >>> f() 3 >>> h.val 0 >>>
    1
  19.  @TechWithTim  no , I get that. Iterables have a dunder iter, iterators have a dunder next. I think this is a subtle python point that needs elucidaiton. I know I've messed it up dozens of times.
    1
  20. I have never seen good code is a .m file Just a wett procedural mess
    1
  21. for grabbing the second element of a list of containers, I kinda like: >>>second = list(map(itertools.itemgetter(2), list_of_containers))
    1
  22. omg, I can help it. for star magic: >>>numbers = print >>>numbers(*lst) I know, it was a toy example.
    1
  23. performance over clarity? wt_?
    1
  24. As a guru, I think Tim’s approach is entirely the logical way to explain python.
    1
  25. Rtfm.
    1
  26. I don't understand the line btw lambda and static methods?
    1
  27.  @aflous  I wrote lambda (single statement) that computed the partitions of an integer, so you can get pretty sick.
    1
  28. what do you mean getters/setter? property and property().setter, yes, but if you're writing methods called set_radius(self, radius) or get_radius(self), then no. Never. It's not java.
    1
  29. omg, I was on a NASA code project and it was full of: space_craft . get_intrument() . get_sensor . get_channel() . get_shelves() . get_shelf() . get_temperature() (and setters like that!) calls, and the 1st thing I did was turn it all in to @property and dunder getattr / init magic, (they were doing bare init class with post attribute setters), oh, and they had one class per module...so they came in Monday freaking out "85% of the code base is gone!" I raised my hand and said "you're welcome". It was pretty rocky after that ;-)
    1
  30. I have a nice context manager that chdir s to a remote directory to do work and write files, and always comes home no matter what. Not all dirs get good data every day
    1
  31. I think all functions should be pure: the result is always the same for the same input, so a parameterless function is useless. Many beginners' OOP is just a script in a class. Tips to avoid that: > A class has one reason to change (see: SRP) > don't do work in dunder init (it's very tempting. Replace with a classmethod helper that does the work and then returns the instance) > make the entire instance in dunder init (pass in all parameters once, no setting later). > only modify instance attributes in its own class methods > avoid functions (on your classes): A class should provide all the methods its instances need. > a method should do one thing (or farm out sub tasks to lesser methods)....e.g., don't have a: def do_all(self):....
    1
  32. yes, self (or whatever you call the 1st parameter in an instance method) is the instance, but the dunder enter/exit methods are called as part of the with...as statement.
    1
  33.  @quitchiboo  right, and 'clever' is not a compliment in python.
    1
  34.  @quitchiboo  clever is unpythonic, as it affects readability: >>>import this
    1
  35. I'm stuck at 3.8 for security, and dataclass is available.
    1
  36. even/odd. since if is the new goto (harmful), I select with: {0: 'even', 1: 'odd'}[x % 2] which is why devs hate me. Actually, I'd map the list to (2)._rmod_ and map that to dict.getitem.
    1
  37.  @nirajraut9408  so it calls new and returns a class, otherwise it calls object.__new__() and then tries to instantiate with __init__. His example is bit too simple, in that it just wraps type to show functionality (which needs to be done), but in a real life use case, you may want to choose a class based on the inputs...and only type can do that. For instance, I have spacecraft telemetry in classes: Current, Voltage, Temperature, and I can tell which one I want by the filename. So I'll do something like: class Telemetry(type): klasses = dict(current=Current, voltage=Voltage, temperature=Temperature) def __new__(cls, name): return cls.klasses[name].fromfile(name) Here fromfile are a class method (constructor helpers) that reads the file (in a context manager), and instantiates the object with values. So in use, I just say: >>>telemetry = Telemetry(filename) and get the data unpacked according to protocol w/o having to worry about picking the right object to put it in. Then I just just say: >>>telemetry.plot() and a house keeping plot is generated, without the code explicitly worrying about what kind of telemetry I need to look at. Note that the whole thing is accomplished with ZERO if/then clauses, reducing cyclomatic complexity.
    1
  38.  @multigladiator384  Other languages! 🤢🤢🤮
    1
  39. the list is a list, meanwhile the generator is a function capable of computing the list.
    1
  40.  @piggy3595  the only thing a method should modify are attributes of the class/instance to which it belongs. So put the data structure in the class and have at it.
    1
  41.  @smaransure2234 yes everything is an object
    1
  42. it's nice to see a demo without annoying type hints clutter, though spaces around the keyword operator hurt.
    1
  43. In physics: Vector are geometric objects that transform under the adjoint representation of the rotation group, SO(3)....which is a symmetry of Euclidean space. In math: Complex numbers, C, are important b/c the fundamental theorem of algebra, which states that all polynomials with complex coefficients have roots in C. ie., it's a closed field. They're also critical in quantum mechanics, as phase plays a huge roll in time evolution, translations, and rotations.
    1
  44. I like lambdas (3:30). The other options is key=operator.attrgetter('age')
    1
  45. Classes make instances as you desire with dunder init. Meta classes make classes as you desire with dunder new
    1
  46. camelCase tells devs you're not a pythonista.
    1
  47. "Learn to Code" is dead.
    1
  48. That is why even experts lint.
    1
  49. Nested list comp: >>>flattened = list(itertools.chain(*matrix)) to quote Raymon Hettinger: "Did you learn something today?"
    1
  50. On tuple type (circa 17:40), what if my tuple is 10,000 ints? Do I need to code: x: Tuple[int, int, ....10000 times...int]? Seems Akward
    1