GHC can be compiled with GHC 5.0 (or something around there). If they add a new feature, they don't use it in GHC for years and years.

*Can* be compiled with GHC 5.0, or *is* compiled? http://haskell.org/ghc/docs/6.8.1/html/users_guide/release-6-8-1.html says that the pointer tagging in 6.8.1 improves most code by 10-15%, "and as a result the compiler is also faster", implying that ghc-6.8.1 is used to compile itself. I'm just curious. -Brent

On Sun, 2007-11-11 at 07:43 -0500, Brent Yorgey wrote:

GHC can be compiled with GHC 5.0 (or something around there). If they add a new feature, they don't use it in GHC for years and years.

*Can* be compiled with GHC 5.0, or *is* compiled?

Can. The version it is compiled with it always itself because it does a two stage bootstrap. I think for 6.8.x the minimum bootstrap version is now 6.0 or 6.2. Duncan

First there was an unnamed language (dynamically typed) that looked a lot like SASL. It was compiled to VAX assembly code and the compiler was written in C. I implemented this language in the summer of 1981. Using this language Thomas Johnsson and I wrote the first LML compiler. Once that was working reasonable we rewrote the LML compiler in LML. This was probably 1982. The LML compiler then got improved over the years. Soon after Haskell 1.0 was presented to the world I did a Haskell front end for th LML compiler, and that's hbc. The hbc compiler was then used to bootstrap ghc. -- Lennart On Nov 11, 2007 3:49 PM, Stefan O'Rear wrote:

On Sun, Nov 11, 2007 at 11:07:29AM +0000, Neil Mitchell wrote:

...

Hi

...

...

"...if GHC is written in Haskell, how the heck did they compile GHC in the first place?"

GHC was not the first Haskell compiler, hbc was the main compiler at some point, so I suspect they used hbc. There was also lazy ML which I suspect was used to bootstrap hbc - but I'm not sure of the details.

Hbc didn't need to be bootstrapped because it isn't written in Haskell - it's written in Lazy ML. Lazy ML would need to be bootstrapped, but since it has (almost?) the same syntax as Standard ML, I suspect the first versions of lml were written in SML. (Can you clarify, L. Augustsson?)

Stefan

-----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.6 (GNU/Linux)

iD8DBQFHNyR9FBz7OZ2P+dIRAtpRAJ9ezJsVyL2PW1J4o+wujmnol6vj0gCdH3xv vcF9yk40pbAKct7mHuJSzfw= =cxUQ -----END PGP SIGNATURE-----

_______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe

Dan Piponi wrote:

This isn't a question specific to ghc.

Most certainly not. ;-) (Well, except that I asked "where did GHC come from", which is pretty GHC-specific.) However, it seems the general point of confusion is that writing (say) a minimally-working C compiler intuitively "seems" quite easy (after all, C is an extremely low-level language), whereas the constructs in Haskell bear no resemblence to any machine-level constructs, and it "seems" unthinkable that you could possibly write such a compiler in anything but Haskell itself.

In general, the process of bootstrapping compilers and porting them to new platforms can be described by T-diagrams. When I did a web search on T-diagrams the first hit I found, http://proglang.informatik.uni-freiburg.de/teaching/compilerbau/2004/T-diagr... seemed to summarise them well.

Ooo... interesting.

Neil Mitchell wrote:

Hi

...

bear no resemblence to any machine-level constructs, and it "seems" unthinkable that you could possibly write such a compiler in anything but Haskell itself.

Hugs is written in C.

Really? :-. Well anyway, I didn't say it can't be done - I said it *looks* like it can't be done. ;-) (Seriously, how big is Hugs? It must be quite large...)

On 12/11/2007, at 4:32 AM, Neil Mitchell wrote:

Hi

...

...

...

bear no resemblence to any machine-level constructs, and it "seems" unthinkable that you could possibly write such a compiler in anything but Haskell itself.

Hugs is written in C.

Really? :-.

Really :-)

...

(Seriously, how big is Hugs? It must be quite large...)

56111 lines, with an additional 5917 for the WinHugs bit.

If I remember correctly, the early versions of the Clean compiler were written in C. Then at some stage they re-wrote it in Clean. Cheers, Bernie.

On 12/11/2007, at 4:08 PM, Michael Vanier wrote:

Bernie Pope wrote:

...

If I remember correctly, the early versions of the Clean compiler were written in C. Then at some stage they re-wrote it in Clean.

You could say they cleaned it up.

It was a dirty job, but now it is self cleaning.

Expressiveness certainly makes it easier, but nothing (other than sanity...) stops you from writing a Haskell compiler in, say, COBOL.

*I* would stop you. Friends don't let friends write in COBOL.

Brent Yorgey wrote:

Expressiveness certainly makes it easier, but nothing (other than sanity...) stops you from writing a Haskell compiler in, say, COBOL.

*I* would stop you. Friends don't let friends write in COBOL.

That's the funniest thing I've read today. You literally just woke my mum up by making my laugh. Thanks! :-D

On Nov 11, 2007, at 17:26 , jerzy.karczmarczuk@info.unicaen.fr wrote:

Andrew Coppin writes:

...

Brent Yorgey wrote:

...

Expressiveness certainly makes it easier, but nothing (other than sanity...) stops you from writing a Haskell compiler in, say, COBOL. *I* would stop you. Friends don't let friends write in COBOL.

That's the funniest thing I've read today. You literally just woke my mum up by making my laugh. Thanks! :-D

I would be the last who wanted to spoil such a good joke. But... tell me please, ANYONE, who takes part in this inspiring exchange: How many COBOL programs have you written in your life? How many programs in Cobol have you actually SEEN?

I've maintained an accounting system and a material requirements planning system written in COBOL. I've written various random things including a Newton's Method program (because I was taking an entry level computing-for-math course which bored me out of my skull --- wasn't allowed to place out of it, and already knew all of it, so I got my revenge by submitting assignments in random languages). COBOL has its place. Compiler design is *not* it. -- brandon s. allbery [solaris,freebsd,perl,pugs,haskell] allbery@kf8nh.com system administrator [openafs,heimdal,too many hats] allbery@ece.cmu.edu electrical and computer engineering, carnegie mellon university KF8NH

G'day all. Quoting jerzy.karczmarczuk@info.unicaen.fr:

But... tell me please, ANYONE, who takes part in this inspiring exchange: How many COBOL programs have you written in your life?

As you well know, only one COBOL program has ever been written. The rest are just modifications of it. Actually, a more interesting problem is what you'd replace COBOL with, and how you'd go about it. Wouldn't it be nice if there was a modern language that you could write or rewrite new parts of your COBOL application in, and it all worked seamlessly with what you already had? Cheers, Andrew Bromage

G'day all. I asked:

Actually, a more interesting problem is what you'd replace COBOL with, and how you'd go about it. Wouldn't it be nice if there was a modern language that you could write or rewrite new parts of your COBOL application in, and it all worked seamlessly with what you already had?

Henning Thielemann replied:

Do you think of something like ABAP? http://en.wikipedia.org/wiki/ABAP

I'd like to respond with some combination of: * I never try to think of ABAP. * No, I said "modern language". But the fact is, I know nothing about ABAP apart from having heard of it once. Not that this has ever stopped us having strong opinions, right? To be a true COBOL replacement, I think that one very important feature is that it is link-compatible with existing COBOL code. You're never going to be able to replace a 6MLOC COBOL monster in any manner other than piecemeal. Cheers, Andrew Bromage

Henning Thielemann:

ajb@spamcop.net wrote:

...

Quoting jerzy.karczmarczuk@info.unicaen.fr:

...

... tell me please: How many COBOL programs have you written in your life?

As you well know, only one COBOL program has ever been written. The rest are just modifications of it.

Actually, a more interesting problem is what you'd replace COBOL with, and how you'd go about it. Wouldn't it be nice if there was a modern language that you could write or rewrite new parts of your COBOL application in, and it all worked seamlessly with what you already had?

Do you think of something like ABAP?

A nice mixture of Cobol, SQL, a Report Writer language... Gents, we can continue for years, not knowing where we are going with this discussion. I must say that I got a little nervous. Two days of speculation on what this awful GHC has been bootstrapped from, as if the fellows who *did it* couldn't be asked directly! They read this list, and laugh... Then a ha_ha_ha over the the corpse of Cobol. OK, perhaps not a corpse... Plenty of people, and some companies, like Fujitsu, still predict a bright future for this language. God bless them. Obviously, a primitive, hierarchic database approach to programming is something very far from compilation issues. But, frankly, if you permit a serious remark: I think that there will be a serious breakthrough in the Haskell popularity in the Great World, when Haskell is applied to the construction of a fabulous, optimized ... ... COBOL compiler. Or Fortran, or SQL, or Basic. Or whatever, all *BUT* Haskell. == I have an anecdote for you. Real one. (An anecdote is *not* a joke). During my studies we had to make a small plastic box for some electronic circuit (it was hundred years ago, don't forget...) We bought some pieces of something related to polymethyl methacrylate (for profans: plexiglas, acrylite, whatever). Awful stuff. Too brittle! But, but, we found for it a fabulous usage! When dissolved in some solvent manufactured from ether + acetone + some other tasty beverage, it transformed into a viscous substance which was a *wonderful*, ideal glue, for gluing these damn pieces together. So, the lesson we learnt was the following: we got a substance whose main application was to make out of it a glue to glue itself. Since the boss of the Lab would not appreciate the high level philosophical issue related to our observation, we had to find some other usage of this glue, so we glued his chair to the floor. After this hundred years they still look for the culprit. Jerzy Karczmarczuk

I have a copy of "COBOL for Dummies" which I bought as a joke and have never dared read. Mike jerzy.karczmarczuk@info.unicaen.fr wrote:

Andrew Coppin writes:

...

Brent Yorgey wrote:

...

Expressiveness certainly makes it easier, but nothing (other than sanity...) stops you from writing a Haskell compiler in, say, COBOL.

*I* would stop you. Friends don't let friends write in COBOL.

That's the funniest thing I've read today. You literally just woke my mum up by making my laugh. Thanks! :-D

I would be the last who wanted to spoil such a good joke. But... tell me please, ANYONE, who takes part in this inspiring exchange: How many COBOL programs have you written in your life? How many programs in Cobol have you actually SEEN? Jerzy Karczmarczuk

_______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe

It's interesting that the article completely fails to mention hbc which I know they used during the GHC bootstrap. Oh well. :) On Nov 11, 2007 2:41 PM, Richard Kelsall wrote:

Andrew Coppin wrote:

...

"...if GHC is written in Haskell, how the heck did they compile GHC in the first place?"

The paper "A History of Haskell: Being Lazy With Class" by Paul Hudak, John Hughes, Simon Peyton Jones and Philip Wadler is a good read.

http://research.microsoft.com/~simonpj/papers/history-of-haskell/index.htm

A snippet

"GHC was begun in January 1989 at the University of Glasgow, as soon as the initial language design was fixed. The first version of GHC was written in LML by Kevin Hammond, and was essentially a new front end to the Chalmers LML compiler. This prototype started to work in June 1989, just as Peyton Jones arrived in Glasgow to join the burgeoning functional programming group there. The prototype compiler implemented essentially all of Haskell 1.0 including views (later removed), type classes, the deriving mechanism, the fullmodule system, and binary I/O as well as both streams and continuations. It was reasonably robust (with occasional spectacular failures), but the larger Haskell prelude stressed the LML prelude mechanism quite badly, and the added complexity of type classes meant the compiler was quite a lot bigger and slower than the base LML compiler. ... GHC proper was begun in the autumn of 1989, by a team consisting initially of Cordelia Hall, Will Partain, and Peyton Jones. It was designed from the ground up as a complete implementation of Haskell in Haskell, bootstrapped via the prototype compiler. The only part that was shared with the prototype was the parser, which at that stage was still written in Yacc and C. The first beta release was on 1 April 1991 (the date was no accident), but it was another 18 months before the first full release (version 0.10) was made in December 1992."

_______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe

No worries. I read the History of Haskell paper twice, but not the final version, I guess. As far as I remember you started the bootstrapping with the ghc built on the LML compiler. But as hbc became available it was soon more reliable and produced better code, so you switched to that until the new ghc could compile itself. That's how I remember it. There's also a bunch of Prelude related code in the ghc libraries that originates with hbc. I've been to lazy to send any "bug" reports about this. (Just to get a line of attribution.) -- Lennart On Nov 12, 2007 9:33 AM, Simon Peyton-Jones wrote:

| It's interesting that the article completely fails to mention hbc | which I know they used during the GHC bootstrap. Oh well. :) | | On Nov 11, 2007 2:41 PM, Richard Kelsall wrote: | > Andrew Coppin wrote: | > > "...if GHC is written in Haskell, how the heck did they compile GHC in | > > the first place?"

Lennart, I'm sure this is my fault, and I do apologise. If we used hbc in the bootstrap, I'd completely forgotten; I thought we'd just used the LML compiler, which is indeed credited. (Perhaps not everyone knows that you wrote the LML compiler too, with Thomas Johnsson, of course. It completely changed the landscape of lazy-functional-language compilers.) What I do remember is that we constantly compared ghc's performance to hbc's, and took ages to catch up!

I wish we'd corrected any errors of fact before the paper was published. (It was on the Web for 6 months before publication to accumulate corrections, but errors and omissions remain our fault.)

Simon

_______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe