DEV Community

Discussion on: Six Things You Thought Senior Devs Did (But We Don't)

 
codemouse92 profile image
Jason C. McDonald

When I look at a language...Python, it's clear to see the harm that piling on features to an existing language can do.

I dunno, Python is my primary language, and I do not see this at all. I, like most of my fellow Pythonistas, love Python with all the passion Rust coders love their language with. I think it's an opinion thing.

Thread Thread
 
jmccabe profile image
John McCabe • Edited

I'm not a big fan of python (nor any language where a slice/range written as 2:5 gives you elements 2, 3 and 4, but NOT 5), but I use it quite a lot for quick and dirty scripting. As far as I can tell, it took the best part of 10 years for Python 3 to overtake Python 2 in usage, almost solely as a result of a rash decision to make 3 substantially incompatible at the source level. I think that did Python a lot of harm. Contrast that to the Ada 83 - > 95 update, whose goal was to provide new features while keeping incompatibilities to a minimum (archive.adaic.com/intro/tech/8395c...).

In my opinion, this highlights how many language authors (I hesitate to call them designers) chuck something out there prematurely with the hit, in terms of language churn, being taken by the users.

Java, in its early days, was a prime example whereby a ton of breaking changes were made on every new revision. Sadly, following a few years of relative stability, Java sounds like it's reverting (admittedly based on what I've read, as opposed to using it) and C++ is getting out of hand with numerous half-baked features going in, with disappointing implementations, and the promise of more 'improvements' in the next revision, 3 years down the line.

Two things I'd been looking forward to in 'modern C++' were enum class and std::array, my hope being that, somehow, they'd manage to implement something almost matching the power and expressiveness of Ada arrays from 35 years earlier. Sadly, no. std::array is pretty much a wrapper round a C array, is almost impossible to create without a constant size, and still uses zero-based indexing at the source code level. enum class doesn't even match Java's enum in usefulness.

In addition to that, I watched part of a video where Herb Sutter was talking about potentially introducing 'interfaces' and asked a couple of audience members what they thought the default visibility of the methods and attributes in the example he showed them would be. When he went on to ask if they thought it would be a good idea if methods were public, by default, and attributes private, they agreed! You have a language that already confuses some people by having two composite type mechanisms, class and struct, with different default visibility of ALL members, and now you want to introduce a 3rd, abstract one, with different visibility depending on whether it's a method or an attribute? They must be out of their tiny minds.

As far as I can see, the people involved in the C++ revisions appear to have lost sight of the fact that bigger and bigger software systems don't need more and more complicated and confusing features and syntax; they need more clarity, simplicity and expressiveness as you can't expect all the people working on these systems to know a language as excessive as C++ inside out.

Sorry - went off track there. Python, yeah, I tend to believe Python is kind of in that mostly stable period; whether it will follow Java and C++ back to instability remains to be seen.

Thread Thread
 
val_baca profile image
Valentin Baca • Edited

I have to strongly disagree on how slices are handled. It's standard for the range to be [x,y), i.e. inclusive of the first index x and exclusive of the second index y. It gives you a consistent # of elements equal to (y-x). So (3,5) should give 2 elements (5-3) starting at 3.

Standard as in used across many languages (links to single examples from each): C++, Java, Javascript, Go

It's as ingrained as the standard for-loop: for (int i = 0; i < n; i++) which is effectively [0,n)

Slices aren't ranges. I see that Ada uses ranges, but they're distinct from ranges.

I also have to disagree that Java is unstable b/c it has features added with new features. I've done a lot of Java migrations, but they're relatively minor. Very few deprecations are breaking and for the most part you can take old code and just run it on the latest JVM.

Thread Thread
 
val_baca profile image
Valentin Baca

I was more referring to GvR leaving the language as leader: mail.python.org/pipermail/python-c...

I'll admit that a majority of my interaction with python has been migrating some ancient (5 years old) scripts from Python 2 to 3 and the experience was not pleasant.

Thread Thread
 
jmccabe profile image
John McCabe • Edited

I'm happy to bow to your superior knowledge of recent Java changes.

As for ranges etc, please check out the use of the range keyword and 'Range attribute in Ada. An array slice in Ada is a subsection of the array defined by a range.

Your suggestion of there being a 'standard' definition of a range in programming appears to be using references to languages that didn't exist when Ada had been using array slices and the range keyword and attribute for a number of years. Also, the suggestion of a 'standard' for loop makes no sense. The specific example you give is common to C, C++ and Java, at least, where array indices have not been abstracted away from machine representations. Even if you go back before Ada became an international standard, in 1983, it had introduced superior array handling that didn't rely on users managing (often badly) the mapping of array indices from the problem domain to the machine domain.

Thread Thread
 
codemouse92 profile image
Jason C. McDonald • Edited

Most of us in the Python world see the transition from GvR's BDFL tenure, as good as he was, to the formal Steering Committee as a good thing. It's certainly brought a lot of improvements to the language, its internal development processes, and the community as a whole.

(Mind you, GvR is still very much involved, just with a layer of removedness that frees him up to pursue new ideas in Python w/ Microsoft funding.)

Python 3 was weird by nature of improving on some mighty weird things in Python 2, that lead to some things needing to be broken. That's why Python 2 wasn't deprecated until Python 3 was on full feature parity.

Thread Thread
 
val_baca profile image
Valentin Baca

That's good to hear. Thanks for your perspective!

Thread Thread
 
jmccabe profile image
John McCabe • Edited

Sorry for the late reply (and possibly in a weird place; I'm not sure Dev's threading implementation is as optimal as it could be) but I've only just found out that Ruby, see ruby-doc.org/core-2.5.1/Range.html, treats a 'range' as Ada does; inclusively, if you use the simple s..e format. An extra dot is required if you want to use the exclusive (i.e. not including the end index) version, s...e. In addition, if you use ::new to create a range, you need to explicitly specify the 3rd, 'exclude_end', parameter if you want the non-default, exclusive, behaviour.