> > If I had to pick someplace to focus effort on improving tooling for Julia > in general, I’d look at improving/adding a network interface to the REPL.
If anyone is interested in working on this, one approach is to implement the server side of the Jupyter protocol in pure Julia. So far the network-integrated-REPL role has been filled by the Jupyter/IPython protocol, benefiting from the years of design iteration there. In many cases this also leverages a broader community of developers who are interested in (I)Python support for a particular editor. The main (and very moderate) disadvantage is the Python dependency, but that could be eliminated. On Mon, Sep 14, 2015 at 8:21 AM, Joshua Ballanco <[email protected]> wrote: > > On September 14, 2015 at 14:17:43, Daniel Carrera ([email protected] > (mailto:[email protected])) wrote: > > On 14 September 2015 at 12:40, J Luis wrote: > > > > > > > > > segunda-feira, 14 de Setembro de 2015 às 09:26:05 UTC+1, Daniel > Carrera escreveu: > > > > > > > > On 14 September 2015 at 08:16, Uwe Fechner wrote: > > > > > While I understand your point, the success of a new programming > language depends on the availability of a good IDE. > > > > > > > > No it doesn't. > > > > > > > > C, C++, Perl, Python, Fortran, JavaScript, PHP, and arguably even > Java became successful long before they acquired an IDE. I think that there > are more languages that became successful without an IDE than with one, so > let's not overstate the issue. An IDE is *good* to have because *some* > people want them. Good documentation is more important. Having the right > features and being at the right place at the right time is even more > important. > > > > > > Yes it does (IMO off course) > > > > > > This is not a matter of opinion. This is an empirical claim. With little > effort we can define a criterion for language success, and determine > whether any language has ever become successful before it acquired an IDE. > A single example (e.g. Fortran) falsifies the claim. In fact, I would make > a stronger statement: that MOST successful languages achieve success before > acquiring an IDE. Off the top of my head, I offer the following successful > languages: > > > > Without IDE: C, C++, Perl, Python, Fortran, JavaScript, PHP, Java > > > > With IDE: C#, VisualBasic, Matlab > > I suppose you could also take the counter-counterpoint of LISP. People not > only built IDEs but entire *machines* tailored specifically to running and > debugging LISP, and it still hasn’t (really) caught on (yet). > > That said, I think the Clojure community provides a useful example for how > to approach the editor/IDE debate. All the early Clojure developers used > emacs, and much of the early community was either on emacs or vim (yes, > there were a few of us). In the intervening 7 or so years, though, as new > developers who were familiar with other IDEs entered the community, they > began projects to develop plugins for their IDE of choice. As such, Clojure > now has first-rate plugins for both Eclipse and IntelliJ. > > It was really only later that projects were started to build “true” > Clojure IDEs, and still I don’t think any of these surpass (or even really > approach) the utility of the IDE plugins (the three IDEs of which I’m aware > are: LightTable, NightCode, and clooj). > > One important element that allowed much of this for Clojure was the early > development of nREPL, the network-enabled REPL. With this, all editors/IDE > plugins stand on equal footing with access to the REPL. I noticed in the > code to REPL.jl there’s a function `start_repl_server`, but it doesn’t seem > to be used anywhere. > > If I had to pick someplace to focus effort on improving tooling for Julia > in general, I’d look at improving/adding a network interface to the REPL. >
