Teal was featured on HN today, and one of the comments was questioning the fact that the documentation states that it “compiles Teal into Lua”:

We need better and more rigorous terms in computing science. This use of the compiler word blurs the meaning of interpreted vs compiled languages.

I was under the assumption that it would generate executable machine code, not Lua source code.

I thought that was worth replying to because it allowed to dispel two misconceptions at once.

First, if we want to be rigorous about computer science terms, calling it “interpreted vs compiled languages” is a misnomer, because being interpreted or compiled is not a property of the language, but of the implementation. There have been things such as a C interpreter and an ahead-of-time compiler for PHP which generates machine code.

But then, we get to the main course, the use of “compiler”.

The definition of compiler has never assumed generating executable machine code. Already in the 1970s, Pascal compilers have generated P-code (a form of “bytecode” in Java parlance), which was then interpreted. In the 1980s, Turbo Pascal produced machine code directly.

I’ve seen the neologism “transpiler” being very frowned upon by the academic programming language community precisely because a compiler is a compiler, no matter the output language — my use of “compiler” there was precisely because of my academic background.

I remember people joking around on Academic PL Twitter jokingly calling it the “t-word” even. I just did a quick Twitter search to see if I could find it, and I found a bunch of references dating from 2014 (though I won’t go linking people’s tweets here). But that shows how out-of-date this blog post is! Academia has pretty much settled on not using the “compiler” vs. “transpiler” definition at all by now.

I don’t mind the term “transpiler” myself if it helps non-academics understand it’s a source-to-source compiler, but then, you don’t see people calling the Nim compiler, which generates C code then compiles it into machine code, a “transpiler”, even though it is a source-to-source compiler.

In the end, “compiler” is the all-encompassing term for a program that takes code in one language and produces code in another, be it high-level or machine language — and yes, that means that pedentically an assembler is a compiler as well (but we don’t want to be pedantic, right? RIGHT?). And since we’re talking assembler, most C compilers do not generate executable machine code either: gcc produces assembly, which is then turned into machine code by gas. So gcc is a source-to-source compiler? Is Turbo Pascal more of a compiler than gcc? I could just as well add an output step in the Teal compiler to produce an executable in the output using the same techniques of the Pascal compilers of the 70s. I don’t think that would make it more or less of a compiler.

As you can see, the distinction of “what is a transpiler” reduces to “what is source code” or “what is a high-level language”, the latter especially having a very fuzzy definition, so in the end my sociological observation on the uses of “transpiler” vs. “compiler” tends to boil down to people’s prejudices of “what makes it a Real, Hardcore Compiler”. But being a “transpiler” or not doesn’t say anything about the project’s “hardcoreness” either — I’m sure the TypeScript compiler which generates JavaScript is a lot more complex than a lot of compilers out there which generate machine code.

André Garzia made a nice blog post called “Lua, a misunderstood language” recently, and unfortunately (but perhaps unsurprisingly) a bulk of HN comments on it was about the age-old 0-based vs. 1-based indexing debate. You see, Lua uses 1-based indexing, and lots of programmers claimed this is unnatural because “every other language out there” uses 0-based indexing.

I’ll brush aside quickly the fact that this is not true — 1-based indexing has a long history, all the way from Fortran, COBOL, Pascal, Ada, Smalltalk, etc. — and I’ll grant that the vast majority of popular languages in the industry nowadays are 0-based. So, let’s avoid the popularity contest and address the claim that 0-based indexing is “inherently better”, or worse, “more natural”.

It really shows how conditioned an entire community can be when they find the statement “given a list x, the first item in x is x[1], the second item in x is x[2]” to be unnatural. :) And in fact this is a somewhat scary thought about groupthink outside of programming even!

I guess I shouldn’t be surprised by groupthink coming from HN, but it was also alarming how a bunch of the HN comments gave nearly identical responses, all linking to the same writing by Dijkstra defending 0-based indexing as inherently better, as an implicit Appeal to Authority. (Well, I did read Dijkstra’s note years ago and wasn’t particularly convinced by it — not the first time I disagree with Dijkstra, by the way — but if we’re giving it extra weight for coming from one of our field’s legends, then the list of 1-based languages above gives me a much longer list of legends who disagree — not to mention standard mathematical notation which is rooted on a much greater history.)

I think that a better thought, instead of trying to defend 1-based indexing, is to try to answer the question “why is 0-based indexing even a thing in programming languages?” — of course, nowadays the number one reason is tradition and familiarity given other popular languages, and I think even proponents of 0-based indexing would agree, in spite of the fact that most of them wouldn’t even notice that they don’t call it a number zero reason. But if the main reason for something is tradition, then it’s important to know how did the tradition start. It wasn’t with Dijkstra.

C is pointed as the popularizer of this style. C’s well-known history points to BCPL by Martin Richards as its predecessor, a language designed to be simple to write a compiler for. One of the simplifications carried over to C: array indexing and pointer offsets were mashed together.

It’s telling how, whenever people go into non-Appeal-to-Authority arguments to defend 0-based indexes (including Dijkstra himself), people start talking about offsets. That’s because offsets are naturally 0-based, being a relative measurement: here + 0 = here; here + 1 meter = 1 meter away from here, and so on. Just like numeric indexes are identifiers for elements of an ordered object, and thus use the 1-based ordinal numbers: the first card in the deck, the second in the deck, etc.

BCPL, back in 1967, made a shortcut and made it so that p[i] (an index) was equal to p + i an offset. C inherited that. And nowadays, all arguments that say that indexes should be 0-based are actually arguments that offsets are 0-based, indexes are offsets, therefore indexes should be 0-based. That’s a circular argument. Even Dijkstra’s argument also starts with the calculation of differences, i.e., doing “pointer arithmetic” (offsets), not indexing.

Nowadays, people just repeat these arguments over and over, because “C won”, and now that tiny compiler-writing shortcut from the 1960s appears in Java, C#, Python, Perl, PHP, JavaScript and so on, even though none of these languages even have pointer arithmetic.

What’s funny to think about is that if instead C had not done that and used 1-based indexing, people today would certainly be claiming how C is superior for providing both 1-based indexing with p[i] and 0-based pointer offsets with p + i. I can easily visualize how people would argue that was the best design because there are always scenarios where one leads to more natural expressions than the other (similar to having both x++ and ++x), and how newcomers getting them mixed up were clearly not suited for the subtleties of low-level programming in C, and should be instead using simpler languages with garbage collection and without 0-based pointer arithmetic.

I ran a small experiment where I ported my handwritten lexer written in Teal (which translates to Lua) to lpeg, then ran the compiler on the largest Teal source I have (itself).

lexer time / total time
LuaJIT+hw   -  36 ms / 291 ms
LuaJIT+lpeg -  40 ms / 325 ms
Lua5.4+hw   - 105 ms / 338 ms
Lua5.4+lpeg -  66 ms / 285 ms

These are average times from multiple runs of tl check tl.tl, done with hyperfine.

The “lexer time” was done by adding an os.exit(0) call right after the lexer pass. The “total time” is the full run, which includes additional lexer passes on some tiny code snippets that are generated on the fly, so the change between the lpeg lexer and the handwritten lexer affects it a bit as well.

Looks like on LuaJIT my handwritten parser beats lpeg, but the rest of the compiler (lots of AST manipulation and recursive walks) seems to run faster on Lua 5.4.2 than it does on LuaJIT 2.1-git.

Then I decided to move the os.exit(0) further, to after the parser step but before the type checking and code generation steps. These are the results:

lexer+parser time
LuaJIT+hw   -  107 ms ± 20 ms
LuaJIT+lpeg -  107 ms ± 21 ms
Lua5.4+hw    - 163 ms ± 3 ms
Lua5.4+lpeg -  120 ms ± 2 ms

The Lua 5.4 numbers I got seem consistent with the lexer-only tests: the same 40 ms gain was observed when switching to lpeg, which tells me the parsing step is taking about 55 ms with the handwritten parser. With LuaJIT, the handwritten parser seems to take about 65 ms — what was interesting though was the variance reported by hyperfine: Lua 5.4 tests gave be a variation in the order of 2-3 ms, and LuaJIT was around 20 ms.

I re-ran the “total time” tests running the full tl check tl.tl to observe the variance, and it was again consistent, with LuaJIT’s performance jitter (no pun intended!) being 6-8x that of Lua 5.4:

total time
LuaJIT+hw   -  299 ms ± 25 ms
LuaJIT+lpeg -  333 ms ± 31 ms
Lua5.4+hw    - 336 ms ± 4 ms
Lua5.4+lpeg -  285 ms ± 4 ms

My conclusion from this experiment is that converting the Teal parser from the handwritten one into lpeg would probably increase performance in Lua 5.4 and decrease the performance variance in LuaJIT (with the bulk of that processing happening in the more performance-reliable lpeg VM — in fact, the variance numbers of the “lexer tests” in lpeg mode are consistent between LuaJIT and Lua 5.4). It’s unclear to me whether an lpeg-based parser will actually run faster or slower than the handwritten one under LuaJIT — possibly the gain or loss will be below the 8-9% margin of variance we observe in LuaJIT benchmarks, so it’s hard to pay much attention to any variability below that range.

Right now, performance of the Teal compiler is not an immediate concern, but I wanted to get a feel of what’s within close reach. The Lua 5.4 + lpeg combo looks promising and the LuaJIT pure-Lua performance is great as usual. For now I’ll keep using the handwritten lexer and parser, if only to avoid a C-based dependency in the core compiler — it’s nice to be able to take the generated tl.lua from the Github repo, drop it into any Lua project, call tl.loader() to register the package loader and get instant Teal support in your require() calls!

Update: Turns out a much faster alternative is to use LuaJIT with the JIT disabled! Here are some rough numbers:

JIT + lpeg: 327ms
5.4: 310ms
JIT: 277ms
5.4 + lpeg: 274ms
JIT w/ http://jit.off: 173ms
JIT w/ http://jit.off + lpeg: 157ms

I have now implemented a function which disables the JIT on LuaJIT and temporarily disables garbage collection (which provides further speed ups) and merged it into Teal. The function was appropriately named:

What the principles that underlie the software you make?

This short story is not really about htop, or about the feature request that I’ll use as an illustration, but about what are the core principles that drive the development of a bit of software, down to every last detail. And by “core principles” I really mean it.

When we develop software, we have to make a million decisions. We’re often driven by some unspoken general principles, ranging from our personal aesthetics on the external visuals, to our sense of what makes a good UX in the product behavior, to things such as “where does bloat cross the line” in the engineering internals. In FOSS, we’re often wearing all these hats at the same time.

We don’t always have it clear in our mind what drives those principles, we often “just know”. There’s no better opportunity to assess those principles than when user feedback asks for a change in the behavior. And there’s no better way to explain to yourself why the change “feels wrong” than to put those principles in writing.

Today was one such opportunity.

I was peeking at the htop issue tracker as an end-user, which is a refreshing experience, having recently retired from this FOSS project I started. I spotted a feature request, asking for a change to make it hide threads by default.

The rationale was sensible:

People casually using htop usually have no idea what userland threads are for.
People who actually need to see them can easily enable them via SHIFT+H.

With them currently enabled by default, it is very inconvenient to go through the list and see what is running, taking up RAM, CPU usage and whatnot, therefore I think it’d be more user-friendly to not show them by default.

He proceeded to show two screenshots: one with the default behavior, full of threads (in green) mixed with the processes, and another with threads disabled.

When one of the current developers said that it’s easier for the user to figure out how to hide things than for them to discover that something hidden can be shown, the counter-argument was also sensible:

Htop can also show Disk IO, which can be arguably very useful, but is hidden by default.

At that point, I decided to put my “original author” hat on to explain what was the intent behind the existing behavior. Here’s what I wrote:

Hi @C0rn3j, I thought I’d drop by and give a bit of historical background as to what was my motivation for showing threads by default.

People casually using htop usually have no idea what userland threads are for.

Yes! I fully sympathize with this sentiment. And the choice for enabling threads and painting them green was deliberate.

You have no idea how many times I was asked “hey, why are some processes green?” over these 15+ years — and no, it wasn’t annoying: each of these times it was an opportunity to teach someone about threads!

(credit: xkcd)

htop was designed to provide a view to what’s going on in the system. If a process is spawning threads like crazy, or if all your cores are overwhelmed because multiple threads of a process are doing work, I think it’s fair to show it to the user, even if they don’t know a thing about threads — or I would say, especially if they don’t know a thing about threads, otherwise these things would be happening and they wouldn’t even know where to look.

Htop can also show Disk IO, which can be arguably very useful, but is hidden by default.

One of my last projects when I was still active in htop development was to add tabs to the interface, so that the user would have a more discoverable path for navigating through these additional columns:

This code is in the next branch of the old repo https://github.com/hishamhm/htop/ — I think the code for tabs is actually finished (though not super tested); it’s pretty nice, you can click on them or cycle with the Tab key, but the Perf Counters support which I was working on, never really got stable to a point of my liking, and that’s when I got busy with other things and drifted away from development — turns out I enjoy writing interface code a lot more than the systems monitoring part!

Anyway, I think that also illustrates the pattern that went into the design: I implemented the entire tabs feature because I wanted to make IO and Perf Counters more discoverable. I wanted to put that information in front of users faces, especially for users who don’t know about these things! I considered the fact that I had implemented the entire functionality of iotop inside htop but people didn’t know about it to be a personal failure, and a practical learning experience: adding systems functionality is useless if the UI design doesn’t get it to users’ hands. (And even I didn’t use the IO features because there was no convenient way of using them.)

I never wanted to implement a tool for “super advanced Linux power users who know what they doing”, otherwise I would have never spent a full line at the bottom showing F-keys, and I would have spent my time implementing Vim bindings (ugh ;) ) instead of mouse support. I’ve always made a point that every setting can be settable via the Setup screen, which you can access via F2-Setup (shown at the bottom line!) and which you can control with the keyboard or mouse, no “edit this config file to use this advanced feature for the initiated only” or even “read the man page” (in fact it only has a man page because Debian developers contributed it!).

I wanted to make a tool that has the power but doesn’t hide it from users, and instead invites users into becoming more powerful. A tool that reaches out its hand and guides them along the way, helping users to educate themselves about what’s happening on the internals of their system, and in this way have more control of it. This is how you hand power to users, not by erecting barriers of initiation or by giving them the bare minimum they need. I reject this dicothomy between “complicated tools for power users” and “stripped-down tools for mere mortals”, the latter being a design paradigm made popular by certain companies and unfortunately copied by so many OSS GUI projects, without realizing what the goals of that paradigm really were, but that’s another rant.

And that’s why threads are enabled by default, and colored in green!

(PS: and to put the point to the proof, I must say that the tabs feature was a much bigger code change than Perf Counters themselves; it included some quite big internal refactors (there’s no “toolkit”, everything is drawn “by hand” by htop) with unfortunately might make it difficult to ressurect that code (or not! who knows?), and of course tabs are user-definable and user-editable!)

The user who proposed the change in the defaults thanked me for the history tidbits and closed the feature request. But in some sense writing this down was probably more enlightening to me than it was for them.

When I started writing htop, I did not set out to create “an instrument for user empowerment” or any such lofty goal. (All I wanted was to have a top that scrolled and was more forgiving with mistypes than circa-2005 top!) But as I proceeded to write the software, every small decision had to come from somewhere, even if done without much deliberate thought, even if at the time it was just “it felt right”. That’s how your principles start to percolate into the project.

Over time, the picture I described in that reply above became clear to me. It helped me build practical guidelines, such as “every setting must be UI-accessible”, it helped me prioritize or even reject feature requests based on how much they aligned to those principles, and it helped me build a sense of purpose for the software I was working on.

If you don’t have it clear to yourself what are the principles that are foundational to the software you’re building, I recommend you to give this exercise a try — try to explain why the things in the software are the way they are. There are always principles, even if they are not conscious to you at the moment.

(PS: It would be awesome if some enterprising soul would dig down the tab support code and ressurrect it for htop 3! I don’t plan to do so myself any time soon, but all the necessary bits and pieces are there!)

Earlier today I quipped on social media: “We need dumb tech and smart users, and not the other way around”.

Later, I clarified that I’m not calling users of “smart” devices dumb. People are smart. The tech should try to not “dumb them down” by acting condescendingly, cutting down on their agency and limiting their opportunities of education.

A few people shared replies to the effect that they wish for smart devices that wouldn’t be at odds with the intents of the user. After all, we all want the convenience of tech, so why settle for “dumb tech”, right?

The question here becomes a game of words: what is a “smart device”, after all?

A technically-minded person will would look at smart devices like smartphones, smart TVs, etc., and say “well, they are really computers”, or “they have computers inside”. But if we want to be technically pedantic, what is a computer? Having a Turing-complete microprocessor running programs? My old trusty microwave for sure has a microprocessor, but it’s definitely not a “smart device”. So it’s not about the internals, it definitely has to do with the end-user perception of the device.

The next reasonable approximation towards a definition is that a smart device allows you to install “apps”. You can extend it with more functionality (which is really making use of the fact that it’s a “computer inside”). Smart TVs and smart phones check this box. However, other home appliances like “smart kettles” don’t, the “smartness” comes from being internet-connected. So, generally, it looks like smart devices are things you either run apps in, or control via apps (from another smart device, of course).

So, allowing for running apps pretty much makes something into a computer, right? After all, a computer is a machine for running software. But it’s really interesting to think what is in fact a computer — where do we draw the line. From an end-user perspective, a game console is also a machine for running software — a particular kind of software, games — but it is not a computer. Is a Smart TV a computer? You can install apps in it. But they are also generally restricted to a certain kind of software: streaming services, video and the like.

Something doesn’t feel like a computer unless you can run any kind of software in it. This universality is a key concept. It’s interesting how we’re slowly driven back to the most fundamental definition of a computer, Alan Turing’s definition of a computer as a universal machine. Back in 1936, before the first actual computer was built during World War II, Turing wrote a philosophy section within a mathematics paper where he made this thought exercise of what it means to compute, and in his example he used the idea of a person doing the computations by hand: reading data, applying rules to process data, producing new data, repeat. Everything that computes follows this model: game consoles, the autopilot in an airplane, PCs, the microcontroller in my microwave. And though Turing had a technical notion of universality in mind, the key point for us here is that in our end-user understanding of a computer and what makes us call some things computers and others not, is that the program (or set of allowed programs) is not fixed, and this is what we see (and Turing saw) as universal: that any program that may be expressed in the computer’s language can be written and run on it, not just a finite set.

Are smart devices universal machines, then, in either sense of the word? Well, you can install new apps in them. Then, it can do new things it couldn’t yesterday. Does that make it a computer? What about game consoles? If I buy a new game (which is effectively new software!), it can also do new things, but you won’t really look at it as a computer. The reason is because you’re restricted on the kind of new software you can make this machine run: it only takes games, it’s not universal, at least from an end-user point of view.

But game consoles are getting “smarter” nowadays! They not only play games; maybe it will also have an app for showing you the weather, maybe it will accept some streaming service… but not others — and here we’re hinting at a key point of what “smart” devices are really like. They are, in fact, on the inside, universal machines that satisfy Turing’s definition. But they are not universal machines for you, the owner. For me, my Nintendo Switch is just a game console. For Nintendo, it is a computer: they can install any kind of software in it in their next software update. They can decide that it can play games, and also access Youtube, but not Netflix. Or they could change their mind tomorrow. From Nintendo’s perspective, the Switch is a universal machine, but not from mine.

The same thing happens, for example, with an iPhone. For Apple, it is a computer: they can run anything on it, the possibilities are endless. From the user’s perspective, it is a phone, into which you can install apps, and in fact choose from a zillion apps in the App Store. But the the possibilities, vast as they may be, are not endless. And that vastness doesn’t help much. From a user perspective, it doesn’t matter how many things you can do with something; what matters are what things you want to do with it, which of those you can and which ones you can’t. Apple still decrees what’s allowed and what isn’t in the App Store, and will also run their own software on the operating system, over which you have zero say. A Chromecast may also be a computer on the inside, with all the necessary networking and video capabilities, but Google has decided that it won’t let me easily play my movies with it, and so it can’t, just like that.

And such is the reality of smart devices. My Samsung TV is my TV, but it is Samsung’s computer. My house is filled, more and more, with computers over which I have no control. And they have control over what I can and what I can’t do with the devices I bought. From planned obsolescence, to collecting data on my habits and selling it, to complicating access to functionality that is there — the common thread in smart devices is that there is someone on the other side controlling the experience. And as we progress towards the “ever smarter”, with AI-based voice assistants being added to more devices, a significant part of that experience becomes the ways it “delights and surprises you“, or, in other words, your lack of control of it.

After all, if it wasn’t surprising, if it did just what you expected and nothing more, it wouldn’t be all that “smart”, right? If you take all the “smartness” away, what remains is a “dumb” device, an empty shell, waiting for you to tell it what to do. Press a button to do the thing, and the thing happens. Don’t press, it doesn’t do the thing. Install new functionality, the new functionality is installed. Schedule it to do the thing, it does when scheduled, like a boring old alarm clock. Use it today, it runs today. Put it away, pick it up to use it ten years from now, it runs ten years from now. No surprise upgrades. No surprises.

“But what about the security upgrades”, you ask? Well, maybe I just wanted to vacuum my living room. Can’t we design devices such that the “online” component isn’t an indispensable, always-on necessity? Of course we can. But then my devices wouldn’t be their computers anymore.

And why do they want our devices to be their computers? It’s not to run their software in it and free-ride on our electricity bill — all these companies more enough computers of their own than we can imagine. It’s about controlling our experience. Once the user has control over which software runs, they make the choices. Once they don’t, the choice is made for them. Whenever behavior that used to be user-controlled becomes automatic in a “smart” (i.e., not explicitly user-dictated) way, that is a way where a choice is taken away from you and driven by someone else. And they might hide choices behind “it was the algorithm”, which gives a semblance of impersonality and deniability, but putting the algorithm in place is a deliberate act.

Taking power away from the user is a deliberate act. Take social networks, for example. You choose who to follow to curate your timeline, but then they say “no, we want our algorithm to choose who to display in your timeline!”. Of course, this is always to “delight” you with a “better experience”; in the case of social networks, a more addictive one, in the name of user engagement. And with the lines between tech conglomerates and smart devices being more and more blurred, the effect is such that this control extends into our lives beyond the glass screens.

In the past, any kind of rant like this about the harmful aspects of any piece of tech could well be responded with a “just don’t use it, then!” In the world of smart devices, the problem is that this is becoming less of an option, as the fabric of our social and professional lives increasingly depends on these networks, and soon enough alternatives will not available. We are still “delighted” by the process, but just like, 15 years in, a smartphone is now just a phone, soon enough a smart kettle will be just a kettle, a smart vacuum will be just a vacuum and we won’t be able to clean our houses unless Amazon says it’s alright to do so. We need to build an alternative future, because I don’t want to go back to using a broom.