All entries for Thursday 27 June 2019

"And then it just clicked

A bit more of the general "philosophy of programming" today, based on a quote I found on the brilliant "C FAQ", currently here and hopefully there to stay. The quote is from Q 18.9b, on learning resources and says:

A word of warning: there is some excellent code out there to learn from, but there is plenty of truly bletcherous code, too. If you find yourself perusing some code which is scintillatingly clear and which accomplishes its task as easily as it ought to (if not more so), do learn everything you can from that code. But if you come across some code that is unmanageably confusing, that seems to be operating with ten bandaged thumbs and boxing gloves on, please do not imagine that that's the way it has to be; if nothing else, walk away from such code having learned only that you're not going to commit any such atrocities yourself.

It's a very good idea to read other people's code when learning - either completely in the wild, or in the form of snippets on sites like Stack Overflow. But always keep in mind that there is some truly terrible code out there, even in commercial packages. It's easy to fall into the trap of thinking that your problem is just "soooo hard" that it just can't be done elegantly, or readably, or even conforming to standards, at all, and mostly that just isn't the case.

Obviously, different people will find different things to be readable, clear, elegant etc. I have an Undergrad degree including Maths, and I rather like the shorthand notation of sums and sets and implication. To some people though, the symbols are simply "all Greek" (pun thoroughly intended). Writing things in a way everybody will find clear is thus often a losing battle. But there is one very very important thing to avoid at all possible costs : the anti-click.

The Click

The 'click' is that feeling of clarity when you jump from grasping the parts of the thing to understanding the whole. I get it a lot with things like Anagram word puzzles, or even song intro quizzes (the kind where you're meant to guess the title). One moment you're seeing a jumble of letters, trying to hold all dozen in your head, and the next you can 'see' the word it must be. Similarly, one moment you're following along the notes and words and the next the key lyric pops into your head and the song is obvious. Optical illusions do it too - and once you've seen it you struggle to "un-see" it again.

This is the 'click' and it's a real asset that you can develop with practice, by reading plenty of code, and writing plenty of code, until you can 'get the gist' of the thing easily.

Symptoms and Diagnosis

It also comes up a lot in debugging - you just 'get the hang' of the way an error pops up, and know what must have happened. That's why reading issue trackers or mailing lists can feel so confusing - the people answering the questions seem to have some mysterious knack for guessing what's wrong from the most un-intuitive errors.

For example, writing in C you may use a function like 'sqrt', and get a 'undefined symbol' error from the linker, which is saying that it can't work out what that function is. So, being logical, you go back and check that you have absolutely definitely included the "math.h" header, which absolutely definitely contains a sqrt function. What did you do wrong? Well probably you forgot to compile using the '-lm' flag, which means "link against the maths library", because that library contains a lot of compiled code (the code isn't in the math.h file). This error can seem very weird, but actually, once you know the cause, it's "obvious".

We had a go at writing a catalogue of programming bugs a while ago, which tried to sort of encapsulate the 'symptoms' of a bug (available here, PDF download). Any bugs, errors or omissions please let us know (rse {@} warwick.ac.uk). Once you see enough bugs, you'll probably start to 'click' and 'just know' what to look for.

The Anti-Click

I mentioned optical illusions up there, and how you can't seem to "unsee" them. Sometimes code has this same sort of thing - an illusory apparent action or cause which is actually not real. This can be awkward and dangerous, because you, and others, will struggle to "unsee" it. It's hard to give an example that isn't really contrived, so I shall fall back onto a classic bit of C: pointer declarations.

In C, you declare pointers using a '*', such as `int * p;`. This is fine. But what about `int * p, q;`? Is q a pointer? No, it's not. Some people argue that the "proper style" will prevent this anti-click, that is, writing that as `int *p, q` and associating closely the * and the p. This is surprisingly hotly debated (e.g. hereand here) with some really tempting anti-clicks (such as here- do read the answers, as the OP there is wrong) but ultimately does need careful thought. You're probably best using whatever you personally find clearest, and changing that only after consideration.

The golden rule - DO NOT write misleading code, docs etc etc. Make it wordy if you have to, but it's not worth the hassle of an obvious, yet wrong, interpretation. As a last resort, just put in a comment explaining what this does not mean, or does not do.

And the flip side - NEVER trust the click - always verify. Always step through line by line and be sure you're right!

Click Immunity

One last thing - there are people for whom some things just never do 'click'. It's very hard to describe, but easy to spot. It's sort of like trying to explain computers to that one relative or quadratic equations to somebody who can't do maths. They're clearly not stupid, but they just don't seem to 'get it'. You explain one thing, like how to print, and it works, but now they need to print from a different program and isn't it just obvious that you do the same thing?

Some people seem to have a knack even beyond that - they don't just "not get it", it's worse! Bug reports display this sometimes - somebody who always gives a lot of information but somehow never includes the vital pieces. They'll post a 100MB log file, but forget to mention which OS. They'll explain exactly what commands they ran but omit some vital parameter.

There's two sides to this - dealing with people who don't click, and dealing when you can't click.

Hell is Other People

Usually you only have one real option to deal with other people who don't get it - guide them through it. Mostly, it just takes longer, not happens never. You probably want to try explaining things different ways, because what feels neat and obvious to you isn't to them. The Socratic style is handy - ask them what they're trying to do, or what they feel they should do, rather than telling.

I just don't GET IT

I expect everybody, sooner or later, finds something that just doesn't click right. You don't get it - it's unintuitive, your gut feeling is always wrong etc. So what do you do? Treat yourself the same as you would somebody else - read different explanations until you find one that fits you. Ask yourself "what am I trying to achieve". Ask yourself "when I ask this question, what might somebody need to answer it" - and if in doubt, ask them! Stop assuming anything, and walk through every step of the problem.

Walk away and do something else - give your brain time to work things through. Never stop asking questions and always learn from the answers. You don't need to 'click' to do something, you just need to watch for wrong intuition. Keep notes - add comments in your code for you to come back to.

Remember - intution is a skill like any other. It can be wrong. And while you might not have a talent for this thing, but with effort, you can make it work.

Heather Ratcliffe : 27 Jun 2019 12:56 |  Tags: Philosophy Programming |  Comments (0) |  Close comments |  Report a problem