Re: [Caml-list] New Ocaml Plug-in for NetBeans

Re: [Caml-list] New Ocaml Plug-in for NetBeans

[hmf]

Hello again,

Jon Harrop wrote:
> On Saturday 26 July 2008 01:24:02 Erik de Castro Lopo wrote:
>> Jon Harrop wrote:
>>> If I might stick my oar in: why don't the OCaml community write an IDE
>>> for OCaml in OCaml using Camlp4 for parsing with throwback and LablGTK
>>> for the GUI?
>> Most people who actually code in Ocaml do so using the best IDE on
>> the planet, Unix. For those people an IDE is a step backwards and
>> hence they have no interest in writing one.
>

I am sceptical of comments such as those of Erik's (no disrespect
intended here). I suspect these people either have not tried using
an IDE or simply haven't made the effort to learn to use the IDE and
take full advantage of it. Which is surprising since mastering Ocaml
requires much effort, and all here seem to agree that the added
productivity of using Ocaml is worth it ;-). In fact mastering
emacs, vi, etc. with all those "modes" also requires a lot of
work. Why should the use of an IDE be any different?


> Graphical throwback of documentation is invaluable for interactive API
> exploration, particularly in the context of GUI programming (I currently
> trawl through ocamlbrowser's useful but very basic interface). A GUI to
> browse and visualize performance profiles is useful (I currently browse
> gprof's output as plain text files using KWrite). A GUI to visualize
> dependencies is useful (I currently lookup the use of "dot" every time I need
> it and the PostScript output is typically mangled by GhostScript).
>
>> So I have an idea; why don't *you* write a cross platform IDE and if
>> it really is better than Unix then people would use it.
>
> I shall see if it is feasible to develop such an application within an OCaml
> Journal article or two. I think it would be both very useful and a very
> instructive educational exercise combining several of OCaml's strengths.

Jon, I would really be interested if you could report back on your
experiences.

>
> However, the resulting program would most likely be difficult to distribute
> due to licensing issues (e.g. if you want to reuse OCaml's typechecker or
> top-level) and could not be a viable commercial product due to the
> limitations of OCaml itself.
>

I am not sure how it was done in OcalIDE but we have full function
signatures (we need only hover above the function). I guess if parsing
is done via another tool this would not be a problem. Could ask
the OcalIDE folks how its done.

Rgrds,
H.F

Re: [Caml-list] New Ocaml Plug-in for NetBeans

[Richard Jones]

On Sat, Jul 26, 2008 at 10:02:46AM +0100, hmf@inescporto.pt wrote:
> In fact mastering emacs, vi, etc. with all those "modes" also
> requires a lot of work.

You miss the point that most Unix programmers already know emacs or vi
intimately.  I use emacs constantly -- even for writing this very
email, and I have been using emacs for nearly 20 years.  No
"mastering" is needed.

I've used IDEs a bit, but never particularly saw the point.  They seem
to just obscure the workings of the build process, and break the
editor, for no tangible gain.  Emacs can already print the type of an
expression when you hover over it, and can autocomplete symbols.  I've
to find an IDE that can edit email.

There are many problems with the Unix build process -- eg. the horrors
of autoconf/automake/libtool, and possibly better integration with
packaging.  Lack of an all-in-one graphical tool is not exactly high
on anyone's list.

Rich.

-- 
Richard Jones
Red Hat

Re: [Caml-list] New Ocaml Plug-in for NetBeans

[Florian Hars]

Richard Jones schrieb:
> I've to find an IDE that can edit email.

Be careful what you wish for:
http://eclipsemail.org/wiki/index.php/Welcome_to_Eclipsemail

- Florian

Re: [Caml-list] New Ocaml Plug-in for NetBeans

[Erik de Castro Lopo]

hmf@inescporto.pt wrote:

> I am sceptical of comments such as those of Erik's (no disrespect
> intended here).

No offence taken.

> I suspect these people either have not tried using
> an IDE or simply haven't made the effort to learn to use the IDE and
> take full advantage of it.

As little as a 18 months ago I spend about 6 months maintaining
100k lines of mixed C++ and C# code using Visual Studio 2005,
the IDE all the IDE fans rave about. It found the Visual Studio
experience amazingly underwhelming; beyond tedious. The GUI hid
details from me that I thought I needed to know and got in my
way when I thought it should stay out of it.

More recently I spend a couple of months using the Adobe Flash
GUI development tools. Again I found this a woefully tedious
exercise and wished for command line tools to replace the stupid
and annoying GUI. The GUI thought it knew how I wanted to format
my Actionscript code better than I did.

The funny thing is that my preferences for command line tools was
something I developed after my first exposure to an IDE. My first
serios coding was done on Borland's Turbo Pascal and Turbo C IDEs
back in the late 1980s.

In the late 1990s I did a lot of FPGA development using the Xilinx
development tools. It was the inadequacy of these tools which forced
me back to Make because my Makefile understood the build process
I wanted to achieve better than than the Xilinx tools. Later on
in my FPGA design career I would do schematic entry of FPGA designs, 
export a Xilinx XNF netlist, convert the XNY netlist to VHDL using
a utility I wrote and then run that VHDL through a simulator.

This was actually a pivotal event for me because I was able to do
better work by breaking free of the IDE which limited what I could
do.

IDEs still limit what I can do. How many IDEs allow for meta
programming; source code compiling to programs which generate
code which gets compiled to create the final program?

How many IDEs cater for more than one language? The thing is I use
lots of lanaguages. At work I work on a number of projects, some
in C, some in C++ and some in Ocaml. Doing it my way, with Linux
as my IDE, means that apart from the compilers, everything else is
the same. Same editor and same build system (make possibly augmented
with the autotools).

How many people who use multiple languages are willing to learn a
different IDE for each language? Eclipse is not the answer either
because however good it might be for Java its not very good for
other langauges.

> Which is surprising since mastering Ocaml
> requires much effort, and all here seem to agree that the added
> productivity of using Ocaml is worth it ;-).

The same can be said for the Unix IDE, but the UNIX IDE is 100
times more flexible and more capable than any other IDE in
existance. I know Make well enough  to whip up a complex make
file in minutes. I am also intimately familair with the automake/
autoconf/libtool set. Since these tools are so flexible they
adapt to my requirements and never force me to work the way they
are designed.

> In fact mastering
> emacs, vi, etc. with all those "modes" also requires a lot of
> work.

I don't like emacs and vi. My editor of choice for the last 13
years has been nedit (Nirvana Editor) which has syntax highlighting
for dozens of languages (and it easy to add new ones or modify
existing ones), regex search/replace and macros. Its configurable
so over the years I have bent it into the shape I  want. The same
goes for my Unix shell.

> Why should the use of an IDE be any different?

Unix is my IDE and I am reasonably certain that I can do more
with my IDE than you can do with yours :-). By more, I mean
more languages, more meta programming, more custom build
options with more languages.

I suspect that a lot of the people who think Ocaml needs an IDE
are people whose primay development platform is windows.

Erik

PS : Here's a nickle kid. Go and buy yourself a real computer :-).
-- 
-----------------------------------------------------------------
Erik de Castro Lopo
-----------------------------------------------------------------
"The earth is degenerating these days. Bribery and corruption abound.
Children no longer mind parents ...and it is evident that the end of
the world is approaching fast." -- Assyrian Tablet Engraved in 2800 B.C.

Re: [Caml-list] New Ocaml Plug-in for NetBeans

[Jon Harrop]

On Saturday 26 July 2008 11:03:12 Erik de Castro Lopo wrote:
> The same can be said for the Unix IDE, but the UNIX IDE is 100
> times more flexible and more capable than any other IDE in
> existance.

Yet we cannot even get basic documentation about potential completions from 
any Unix development environment for OCaml.

> I know Make well enough to whip up a complex make file in minutes.

Yet Make is not expressive enough so we have OMake, OCamlBuild...

> I am also intimately familair with the automake/ 
> autoconf/libtool set. Since these tools are so flexible they
> adapt to my requirements and never force me to work the way they
> are designed.

That's great but it is the writing of OCaml code that is unnecessarily 
cumbersome, not the building of it.

> > In fact mastering
> > emacs, vi, etc. with all those "modes" also requires a lot of
> > work.
>
> I don't like emacs and vi. My editor of choice for the last 13
> years has been nedit (Nirvana Editor) which has syntax highlighting
> for dozens of languages (and it easy to add new ones or modify
> existing ones), regex search/replace and macros. Its configurable
> so over the years I have bent it into the shape I  want. The same
> goes for my Unix shell.

I assume nedit does not even have basic type throwback, let alone 
documentation throwback?

> I suspect that a lot of the people who think Ocaml needs an IDE
> are people whose primay development platform is windows.

Diversifying to Windows has certainly shown me just how far behind Unix is in 
terms of usability, productivity and modern computing environments like GUIs.

Here is an example: programming the GUI Sudoku solvers in OCaml and F# for the 
OCaml and F#.NET Journal articles. I had years of experience with OCaml but 
little experience of LablGTK2 (I did not know its API at all). I had little 
experience with F# and none with Windows Forms. Yet I wrote the F# 
implementation 10x faster because the Visual Studio mode makes it trivial to 
explore unfamiliar APIs with complete graphical throwback of documentation. 
In contrast, developing the OCaml required me to use "grep" to search the 
LablGTK2 source code distribution from the command line and ocamlbrowser to 
find definitions (but there is no way to jump to related definitions and no 
way to jump back to previous definitions). That is unbelievably tedious in 
comparison.

Provided you only want to write programs that manipulate text and maybe do 
some custom OpenGL, OCaml is awesome. But if you want to write even the most 
mundane GUI application, OCaml is a world of pain compared to the 
alternatives. This could be solved by a decent graphical development 
environment.

Mathematica has by far the best GUI I have ever seen. I think it would be 
fantastic to have such an interface available for OCaml but, as I say, the 
front-end requires tight bindings to the compiler and top-level which is not 
easy with OCaml.

-- 
Dr Jon D Harrop, Flying Frog Consultancy Ltd.
http://www.ffconsultancy.com/products/?e

Re: [Caml-list] New Ocaml Plug-in for NetBeans

[Erik de Castro Lopo]

Jon Harrop wrote:

> On Saturday 26 July 2008 11:03:12 Erik de Castro Lopo wrote:
> > The same can be said for the Unix IDE, but the UNIX IDE is 100
> > times more flexible and more capable than any other IDE in
> > existance.
> 
> Yet we cannot even get basic documentation about potential completions from 
> any Unix development environment for OCaml.

Completions annoyed me immensely in visual studio and the Adobe's
products.

> > I know Make well enough to whip up a complex make file in minutes.
> 
> Yet Make is not expressive enough so we have OMake, OCamlBuild.

I find Make expressive enough. I don't use these others.

> I assume nedit does not even have basic type throwback, let alone 
> documentation throwback?

There are ways of getting this to work with nedit but I never
bothered because I don't like it.

If I need documentation I read the mli files. With bash command
line completion in an xterm I can find the one I want in a second
and I keep it open in a nedit window and then alt-tab between the
window I'm editing and the mli file I'm reading. Sometimes I stick
the two side by side.

> Diversifying to Windows has certainly shown me just how far behind Unix is in 
> terms of usability, productivity and modern computing environments like GUIs.

You sound like someone who never mastered Unix, someone who never
realised that each user needs to mould Unix to their needs rather
than accepting what Unix provides as a default.

Erik
-- 
-----------------------------------------------------------------
Erik de Castro Lopo
-----------------------------------------------------------------
"PHP is a minor evil perpetrated and created by incompetent amateurs, whereas
Perl is a great and insidious evil perpetrated by skilled but perverted
professionals." -- Jon Ribbens

Re: [Caml-list] New Ocaml Plug-in for NetBeans

[Jon Harrop]

On Saturday 26 July 2008 13:07:55 Erik de Castro Lopo wrote:
> Jon Harrop wrote:
> > Yet Make is not expressive enough so we have OMake, OCamlBuild.
>
> I find Make expressive enough. I don't use these others.

The others are particularly useful when you have multiple stages of 
compilation that introduce new dependencies at compile time.

> > I assume nedit does not even have basic type throwback, let alone
> > documentation throwback?
>
> There are ways of getting this to work with nedit but I never
> bothered because I don't like it.

I find type throwback in Emacs invaluable and I miss documentation throwback 
enormously.

> If I need documentation I read the mli files. With bash command
> line completion in an xterm I can find the one I want in a second
> and I keep it open in a nedit window and then alt-tab between the
> window I'm editing and the mli file I'm reading. Sometimes I stick
> the two side by side.

That is exactly what I do when writing OCaml at the moment and I find that it 
leaves a lot to be desired.

For example, I cannot even jump to the definition of an identifier reliably. I 
can look at the identifier and guess where it came from, potentially having 
to manually trawl through directories of source files exactly as you 
describe, hoping to find the correct location from many identifiers with the 
same name. But it would be much easier if I could simply jump directly to the 
location of the definition and then jump back. That could be done from plain 
text editors.

-- 
Dr Jon D Harrop, Flying Frog Consultancy Ltd.
http://www.ffconsultancy.com/products/?e

Re: [Caml-list] New Ocaml Plug-in for NetBeans

[Damien Doligez]

On 2008-07-26, at 17:22, Jon Harrop wrote:

> For example, I cannot even jump to the definition of an identifier  
> reliably.

OCaml 3.11 has extended .annot files that will allow external tools
to do that.  Also, it tells you which function calls are tail calls
and which are normal calls.

-- Damien

Re: [Caml-list] New Ocaml Plug-in for NetBeans

[Lukasz Stafiniak]

On Tue, Jul 29, 2008 at 4:16 PM, Damien Doligez <damien.doligez@inria.fr> wrote:
>
> OCaml 3.11 has extended .annot files that will allow external tools
> to do that.  Also, it tells you which function calls are tail calls
> and which are normal calls.
>
Cool! Are the http://osp.janestcapital.com/files/ocamlwizard.pdf
project participants following this? Would be nice to hear their
progress report :)

Re: [Caml-list] New Ocaml Plug-in for NetBeans

[Jean-Christophe Filliâtre]

Lukasz Stafiniak a écrit :
> On Tue, Jul 29, 2008 at 4:16 PM, Damien Doligez <damien.doligez@inria.fr> wrote:
>> OCaml 3.11 has extended .annot files that will allow external tools
>> to do that.  Also, it tells you which function calls are tail calls
>> and which are normal calls.
>>
> Cool! Are the http://osp.janestcapital.com/files/ocamlwizard.pdf
> project participants following this? Would be nice to hear their
> progress report :)

They tried, indeed (I'm kind of helping in that projet, so I'm aware of
the progress). Unfortunately, even with the CVS version of Ocaml, the
.annot files appear to lack some information. But the solution currently
followed by Ocamlwizard is along the lines of .annot files, and may even
rely on these files in future version of Ocaml.

-- 
Jean-Christophe

[off-topic] was Re: [Caml-list] New Ocaml Plug-in for NetBeans

[Richard Jones]

On Sat, Jul 26, 2008 at 12:40:10PM +0100, Jon Harrop wrote:
> Yet Make is not expressive enough so we have OMake, OCamlBuild...

Make is perfectly expressive enough.

When you start an OCaml project, you certainly need to know a bunch of
stuff to write the autoconf/make framework, and it's not very well
documented.  Almost everyone starts from an existing project -- I
suggest starting from here[1].  IDEs let you start a project much more
easily because they write the boilerplate.

Ah but here's the problem: the boilerplate is meaningful, and sooner
or later you'll need to change it (eg. your project has some complex
code generation or you want to script some automated tests).  Now your
IDE is getting in the way, your beginner has to face all that "stuff"
which was hidden behind the scenes, and (in one IDE I used) you
couldn't edit the boilerplate at all!

Not to mention serious real world problems like collaborating with
people who don't want to use the IDE, version control, cross-
compiling, applying patches, making tarballs & RPMs, uploading to your
website, feeding patches back upstream, integration with l10n tools,
etc.  Most of which are way beyond what IDEs offer.

If you think the good people who develop libvirt could do it using an
IDE, you really don't understand the scope of the problem:

http://git.et.redhat.com/?p=libvirt.git;a=tree
http://git.et.redhat.com/?p=libvirt.git;a=blob;f=configure.in;h=8e04f14131cf68de6eee6eadd05c5704ea8a5d41;hb=HEAD
http://git.et.redhat.com/?p=libvirt.git;a=blob;f=Makefile.am;h=b5082d6a7eaf7c746c3e52d61f6eb952df79db42;hb=HEAD

Rich.

[1] http://hg.et.redhat.com/virt/applications/virt-top--devel click 'manifest'

-- 
Richard Jones
Red Hat

Re: [off-topic] was Re: [Caml-list] New Ocaml Plug-in for NetBeans

[Jon Harrop]

On Saturday 26 July 2008 13:17:55 Richard Jones wrote:
> Ah but here's the problem: the boilerplate is meaningful, and sooner
> or later you'll need to change it (eg. your project has some complex
> code generation or you want to script some automated tests).  Now your
> IDE is getting in the way, your beginner has to face all that "stuff"
> which was hidden behind the scenes, and (in one IDE I used) you
> couldn't edit the boilerplate at all!

Yes. So it would be a good idea to make the IDE as uninvasive as possible. 
Presumably it could even interact with findlib in order to let you choose 
library dependencies graphically.

> Not to mention serious real world problems like collaborating with
> people who don't want to use the IDE, version control, cross-
> compiling, applying patches, making tarballs & RPMs, uploading to your
> website, feeding patches back upstream, integration with l10n tools,
> etc.  Most of which are way beyond what IDEs offer.

Sure. So you continue to do all of those things conventionally. If you really 
wanted you could augment an IDE with functionality to upload to your website 
etc. but I'm not sure that would be particularly beneficial. My main interest 
was better source navigation.

-- 
Dr Jon D Harrop, Flying Frog Consultancy Ltd.
http://www.ffconsultancy.com/products/?e

Re: [Caml-list] New Ocaml Plug-in for NetBeans

[Nathaniel Gray]

On Sat, Jul 26, 2008 at 4:40 AM, Jon Harrop <jon@ffconsultancy.com> wrote:
> On Saturday 26 July 2008 11:03:12 Erik de Castro Lopo wrote:
>>
>> I don't like emacs and vi. My editor of choice for the last 13
>> years has been nedit (Nirvana Editor) which has syntax highlighting
>> for dozens of languages (and it easy to add new ones or modify
>> existing ones), regex search/replace and macros. Its configurable
>> so over the years I have bent it into the shape I  want. The same
>> goes for my Unix shell.
>
> I assume nedit does not even have basic type throwback, let alone
> documentation throwback?

As the guy who implemented one generic form of "throwback" for NEdit,
namely calltips, I can tell you that it certainly does support it and
I've used it extensively.  It's an editor that is extremely flexible
without being *too* obscure, so you can get it to do just about
anything you want.  I have been unable to find another editor that has
the same combination of power, flexibility, and ease-of-use, and I've
looked far and wide.

Having said that, I'm afraid NEdit development has sunk into a tar
pit.  There is no effective leadership on the project and it's
inextricably tied to the Motif toolkit, which means very few new
developers will sign on.  I've given up and switched to jEdit, which
has a similar spirit (platform-independent, language-independent,
flexible) but, being written in Java, a much higher bloat factor.  But
hey, with 2GB of ram it doesn't feel so bad to give 150MB to my text
editor any more.  I would *love* to have an alternative written in
OCaml, since my forays into the jEdit code have left me with
unpleasant feelings...

Cheers,
-n8

-- 
>>>-- Nathaniel Gray -- Caltech Computer Science ------>
>>>-- Mojave Project -- http://mojave.cs.caltech.edu -->

Re: [Caml-list] New Ocaml Plug-in for NetBeans

[Jon Harrop]

On Saturday 26 July 2008 10:02:46 you wrote:
> Jon Harrop wrote:
> > I shall see if it is feasible to develop such an application within an
> > OCaml Journal article or two. I think it would be both very useful and a
> > very instructive educational exercise combining several of OCaml's
> > strengths.
>
> Jon, I would really be interested if you could report back on your
> experiences.

Will do.

> > However, the resulting program would most likely be difficult to
> > distribute due to licensing issues (e.g. if you want to reuse OCaml's
> > typechecker or top-level) and could not be a viable commercial product
> > due to the limitations of OCaml itself.
>
> I am not sure how it was done in OcalIDE but we have full function
> signatures (we need only hover above the function).

Function signatures is one thing but the ocamldoc comments are at least as 
useful as well.

> I guess if parsing is done via another tool this would not be a problem.
> Could ask the OcalIDE folks how its done.

I believe they rewrote everything in Java which, while admirable, certainly 
does not make me want to dive into the source...

-- 
Dr Jon D Harrop, Flying Frog Consultancy Ltd.
http://www.ffconsultancy.com/products/?e