- Haskell-Cafe - Haskell.org

Compilers written in Haskell that generate assembly
by Vanessa McHale 17 Dec '20

17 Dec '20

Hi all, Have any of you written compilers or backends that generate assembly? Lots of projects seem to "end" at LLVM. I have a project (https://hackage.haskell.org/package/kempe) where I used something like maximal munch and a linear register allocator, but I was wondering if there was anything more sophisticated out there! So far I've found http://hackage.haskell.org/package/linearscan-hoopl also Cheers, Vanessa McHale

6 6

Question on optimization
by Todd Wilson 17 Dec '20

17 Dec '20

Dear Cafe, In code of the form case as ++ bs of [] -> ... x:xs -> ... under what circumstances is the overhead of cons-cell creation for the elements of `as` that would happen in an eager language avoided? Always? Is the answer the same if `f` is defined by simple list recursion and we call `f (as ++ bs)`? What if `f` is defined in a different module (or comes from the Prelude)? I know that questions of this type can be answered in specific cases by looking at the (intermediate) code generated by the compiler, but more generally, are there some succinct principles that everyday programmers can hang on their cubicle walls that summarize a lot of what they can expect from the compiler in terms of optimizations? Todd Wilson

2 4

"Natural" polymorphism for n*(n+1)/2
by M Douglas McIlroy 16 Dec '20

16 Dec '20

Some nominally rational functions, e.g n*(n+1)/2, yield integer values for integer arguments. I seek either a way to wrap such a function so it has type Num a => a->a or a convincing argument that it can't be done. Doug

7 15

Teaching High-School one-semester FP (using Haskell)
by Zachi Baharav 16 Dec '20

16 Dec '20

Dear mighty Haskell list, I am a high-school teacher in the US (California, Palo Alto). We have an advanced CS course that students can take after completing the AP-CS-A (in Java). This is usually Juniors and Seniors, after at least 2 or 3 courses of coding (Python+Java). Traditionally, I've done a potpourri of subjects, mostly search algorithms as applied to 'games'. Of course a little bit of data-structure comes in, and each year we did some 'other' interesting subject (like halftoning, barcodes, and so on) and investigated and implemented algorithms there. THIS year, for the second half, I would like us to do Functional-Programming. From what I have seen, my students over the years tend to write more and more convoluted codes, electing first to search on stackOverFlow for some similar pattern, rather than think and find a concise and clean solution. I think FP would supply them with a new way of thinking, which will help with whatever they will write later on. I've been coding with Haskell for about 7 years, so feel ok leading the class. The question: I have many books on Haskell (i think 'all', but who knows. Haskell, Real world Haskell, Learn you a haskell for great good, programming in haskell, and many more). ---> I am looking for something more hands-on and 'fun' for HS students. may i say the buzzword 'project based', or maybe better 'problem based' for our case. Something that I could teach a little, and then we can solve a bunch of problems, and teach a little more, and so on. It doesn't have to be a Book!! Just an outline of a course someone did with associated bunch of problems would be awesome. We have 18 weeks in a semester. My thought right now (if I don't find anything) is to relyon Euler project early problems. These are often clean and simple in Haskell. Or otherwise some CSES problems (which we've done in Java). Ok, long enough email. If you have material and can share here, great! If you have material and want to reach out privately, please do ( zbaharav(a)kehillah.org) If you have a good pointer, that would be appreciated. Thanks in advance for any help! Zachi (Dr. Zachi Baharav, HS teacher (after 20 years in Industry and academia))

11 14

Psa : perhaps time for maintainers to start migrating off Travis ci
by Carter Schonwald 15 Dec '20

15 Dec '20

Hey everyone: it looks like, from my perspective and experiences, that Travis ci should perhaps now be viewed as not open source friendly. Or even converging on hostile? 1) crazy long queue times/ latency for oss ci actions to run 2) very low concurrency on oss builds. 3) very low build build minute caps for oss that require high touch customer support contact to adjust. I’ve started moving my own projects slowly to gh actions for now, though there’s also gitlab ci , src hut and other options that may suit different folks. There’s definitely some ways to keep on having the clever cabal caching we know and love that folks like the Haskell-ci folks and others have hacked out for Travis be available on other platforms, though I don’t think there’s consolidated docs for those yet ? Def seen it discussed though. https://github.com/haskell-CI/haskell-ci/issues/411 Heres a url to my dupe ticket where I share an example naive use of the setup Haskell gh actions Config, definitely not perfect. But kinda amazing to have Mac and Linux and windows ci all in one ! :)

7 14

error: ghcide must be compiled with same GHC version as the project
by Dennis Raddle 14 Dec '20

14 Dec '20

I'm trying to use the Haskell extension with VS Code which, is my understanding, uses ghcide. When I built ghcide, I had the resolver in the stack.yaml set to "nightly-2020-09-02". Associated with this is GHC 8.10.2. When I first created my project I had the resolver set to something like lts-14.20. Associated with this is GHC something like 8.6.5. Later I changed my project resolver to nightly-2020-09-02, same as ghcide. But I'm still getting this error. Perhaps I haven't properly rebuilt my project using the new resolver? Dennis

1 0

Re: [Haskell-cafe] Teaching High-School one-semester FP (using Haskell)
by Johannes Waldmann 13 Dec '20

13 Dec '20

Hi. Teaching is a topic that I do care about, and spend a lot of time on. I want to underline two points that Chris Smith made in his message about the ideas behind "Code World". I am speaking from current experience in teaching an "Intro to CS" course for non-CS students ("Informatik" on https://imweb.imn.htwk-leipzig.de/~waldmann/lehre.html) > easy graphics and UI library with a purely functional style. > ... lets your students experience declarative programming > and build cooI things easily, without needing to > get bogged down in stuff like I/O [...] etc. So much this! Have some basic pictures (rectangle, circle), combinators "row" (besides) and "column" (above), with some basic transformations. This will get students started with the algebraic method (description in the term algebra, evaluation in the domain algebra). I gave homework "produce pictures that resemble letters, traffic signs, etc." Does not need fancy graphics at all - I looked at the "diagrams" package but then decided to go with rectangular black/white pictures only. These restrictions make the "draw a sign" exercise more interesting. And after all, CS is about abstraction. It also allows me to connect to Boolean operations (combine pictures of identical sizes with point-wise Nand, etc.) that are treated in the lecture as well. > get bogged down in stuff like [...] recursion, etc. Again, yes! For my current course, I indeed think that primitive recursion (that is, iteration) is quite enough. E.g., students can draw a chess board from a single white square by flipping the color, and iterated composition. (Of course a CS major should know about the halting problem - but only to avoid it when programming. A non-CS major shouldn't write programs at all.) Advertisment: So I made an experimental "sub-Haskell" language (with online interpreter, integrated into an auto-grading system) (source currently at https://gitlab.imn.htwk-leipzig.de/autotool/all0/-/tree/661-programmable-gr… - I should break it out) with drastically restricted features * just one type (Pictures) * "let" but it's not "letrec" - definitions cannot be recursive * functions but they must be named (no lambdas) * strict evaluation (except for if-then-else which is hard-coded) * just one hard-wired higher order functions: iterate :: (a -> a) -> Nat -> a -> a * well, here I need Nat, and for if-then-else I need Bool, so that's the other two domain types besides Picture * formal parameters of functions must have a type declaration * extensive tracing of evaluation * and check(For) for property-based testing * code with with holes, where students have to fill in the holes, but cannot change anything outside. I can switch off some features (e.g., no functions, or, just unary) and I can switch the underlying domain: instead of pictures, can use sequences (it's not too much of a difference, a sequence is just a one-dimensional picture). I will use this for showing (one method of) compression: find a short straight-line program that produces a given string. Also I was thinking of introducing the "music" domain (scores/performances as in Haskore/Euterpea) but I am undecided about the best way of rendering MIDI in the browser (without 3rd-party JS, see below). Perhaps it doesn't have to be rendered (just print the events) as CS is about abstraction ... So while I try to stay in a Haskell subset (syntactically and semantically) the goal is not to teach some concrete programming language but algebraic/structural thinking. That is, mathematical thinking. Not "computational thinking" - because there is no such thing. If it works, it *is* mathematical... - J.W. NB: another necessary goal of an "intro to all things computer/computing" course is to make students aware of the surveillance economy. Fits nicely in the "computer networks" chapter. > ... get your students started quickly and easily in class, > without waiting for them to install GHC, Cabal/Stack, etc. Yes. That's hard even for CS majors. Well and for me as well. I normally provide software installed in the computer pool, but .. cf. https://github.com/haskell/cabal/issues/7172 Of course "the browser is the OS" is what Evil Corp wants students and teachers to think ..

1 0

Re: [Haskell-cafe] What features should an (fictitious) IDE for Haskell have?
by Luke Lau 11 Dec '20

11 Dec '20

haskell-language-server recently got some basic integration with the retrie plugin, and can offer to do folding and unfolding: https://github.com/haskell/haskell-language-server/pull/266 Luke Lau

2 1

Re: [Haskell-cafe] Teaching High-School one-semester FP (using Haskell)
by Andrew Butterfield 11 Dec '20

11 Dec '20

Dear all > On 11 Dec 2020, at 00:07, Henning Thielemann <lemming(a)henning-thielemann.de> wrote: .... stuff elided > So recursion looks to me as funtional programming's GOTO, and if it is tail recursive, it actually is one. Yes, but it's "GOTO considered beneficial" ;-) Thanks to Zachi for asking the original question - I've already seen lots more ideas for me to chase up for my course! Regards, and Happy Holidays, Andrew > _______________________________________________ > Haskell-Cafe mailing list > To (un)subscribe, modify options or view archives go to: > http://mail.haskell.org/cgi-bin/mailman/listinfo/haskell-cafe > Only members subscribed via the mailman list are allowed to post. -------------------------------------------------------------------- Andrew Butterfield Tel: +353-1-896-2517 Fax: +353-1-677-2204 Lero@TCD, Head of Software Foundations & Verification Research Group School of Computer Science and Statistics, Room G.39, O'Reilly Institute, Trinity College, University of Dublin http://www.scss.tcd.ie/Andrew.Butterfield/ --------------------------------------------------------------------

1 0

Re: [Haskell-cafe] What features should an (fictitious) IDE for Haskell have?
by Olaf Klinke 08 Dec '20

08 Dec '20

On 20-12-07 18:22, André Popovitch wrote: > One feature I've always wanted would be able to "manually inline" a > function. For example, if I have > > foo a = a + a > bar = foo 3 > > I'd like to be able to right-click on the invocation of `foo` and > select > some option to inline it, rewriting the file to read > > foo a = a + a > bar = 3 + 3 > > I think this would make it simple to dive into an abstraction to see > what's > really going on. For example, imagine a new user is confused by this > expression: > > Just 3 >> Just 4 > > They could right-click on `>>` to see that it is identical to: > > case Just 3 of > Just _ -> Just 4 > Nothing -> Nothing > > > Which would allow them to easily see that this expression will always > evaluate to "Just 4" (and hopefully simplify it). Enlightened, they > could > then hit ctrl-z to revert the code to its original state. > > Haskell's purity and laziness makes it one of the only mainstream > languages > where equational reasoning is really possible, so this feature would > play > on Haskell's strengths and provide a feature that couldn't be > replicated in > other languages' IDEs. It would also allow new users to more easily > dive in > and understand certain abstractions (although how it would work in > many > cases would be hard to determine). +1 for this. It is like a step-by-step evaluation/debugger but limited to function calls at the syntactic level. I can image so many past situations where this would have prevented lots of head-scratching. This hypothetical IDE feature might go well with a tree view where inlining a function application is equivalent to expanding a level in a tree (within the same source line number), like so: Folded: [+] button appears on mouse over "foo". bar = foo[+] 3 Expanded: One code line now occupies two lines in the IDE window. [-] button reverts code to previous state. bar = foo 3 [-] 3 + 3 I've seen similar things in action in a demo implementation of [1,2]. Also great for teaching, I guess. Olaf [1] https://www.cs.bham.ac.uk/~pbl/papers/functionalexplainsupp.pdf [2] https://dl.acm.org/doi/10.1145/3110258

1 0