@grammarware

Selected work-related tweets

4 Nov, 16:09 @grammarware: The main problem is not what to teach, but how to set goals. Do we want MSc CS to be coders? Developers? Designers? Mathematicians? Hackers? ※ @
- 16:10 @bigballofmud: @grammarware All of the above except for mathematicans. ※ @
  - 16:13 @grammarware: @bigballofmud the "all of the above" answer needs priorities attached. Why not mathematicians? Algorithm theory, type systems, verification… ※ @
    - 16:18 @bigballofmud: @grammarware That WAS a a priority call. MsC progs already place too little emphasis on dev skills and too much on theory. Then we get 'em. ※ @
      - 16:23 @grammarware: @bigballofmud do we really need universities to train coders/developers? 5 years and a gang of professors is an overkill if you want that. ※ @
        16:52 @bigballofmud: @grammarware Your categories are laden w/implicit caste distinctions I failed to discern at first. If anything, grads 2day are undertrained. ※ @
        5 Nov, 14:13 @grammarware: @bigballofmud that is, "undertrained in what you think they should be trained". Hence, wrong curriculum for the lack of clarity in the goals ※ @
16:08 @grammarware: The never ending "what to teach in CS" holy war breaks down again: joelonsoftware.com stochasticgeometry.wordpress.com timcowlishaw.co.uk ※ @
2:55 @grammarware: Ideas list for the EDD-based tool includes generation of DCGs with XSLT, normalisation of EBNF and other similarly crazy and wonderful items ※ @

3 Nov, 19:49 @grammarware: Sketched the schema for the new language (see last twit) called EDD. Going out for dinner, will hack up a tool & tweet more when I get back. ※ @
19:16 @grammarware: Wearing an "I'm a software language engineer" t-shirt is dangerous: came up with a new language while listening to @notquiteabba's lecture. ※ @

2 Nov, 11:40 @grammarware: Found out syntax highlighting is not always useful: dark red is invisible on a beamer, so 'cat x' worked better at today's lab than 'vi x'. ※ @

29 Oct, 22:40 @grammarware: No such thing as perfection in this world! In one of the very last passes noticed a typo on the first line of the first page of the paper. ※ @

28 Oct, 16:20 @grammarware: 1st lecture on Paradigms went well, lots of students for @notquiteabba to scare away. Let's see how many will show up to my lab next Monday. ※ @
- 17:26 @ashalynd: @grammarware so, on top of other things, you enjoy making the students' life difficult by forcing them to study something ;-) ※ @

23 Oct, 2:50 @grammarware: Disciplined coevolution of tools and papers should be a research domain. #i4cs ※ @

22 Oct, 6:32 @grammarware: Basically, xbgf:flatten is a multiple-production version of xbgf:unchain. We stick to our notion of horisontalisation we had for SLPS. ※ @
6:30 @grammarware: Let the tentative name for the "horizontalize-prime" grammar transformation proposed yesterday by @tvdstorm for now be "flatten". #xbgf ※ @

5 Oct, 23:58 @grammarware: A little hacking moment in the middle of text editing: disambiguated the SDF grammar for BNF, can now parse DMS grammars in MetaEnvironment. ※ @

10 Sep, 20:25 @JBezivin: RT @grammarware : Domain-specific software like that used by any expert group, should be called domainware. ※ @
20:21 @grammarware: Domain-specific software like that used only by pro musicians, or by translators, or by any other expert group, should be called domainware. ※ @

2 Sep, 16:36 @grammarware: Ex-chapter 5 is now Chapter 7, ex-7 is 8, ex-4 is 5, ex-3 is 4, part of ex-2 is 3. Chapters 1 and 6 are lucky to remain in place. #thesis ※ @

24 Aug, 3:20 @grammarware: Camel casing XLDF done & committed to SLPS. Thank God I could automate even this momentary migration, can sleep well knowing it’s error-free ※ @
1:58 @grammarware: Is it wise to change one of the fundamental transformation languages in the middle of the night? 'Cause I’m doing it. #i4cs ※ @

13 Aug, 21:57 @grammarware: Regarding last tweet about high fidelity source code transformation — prooflink: dx.doi.org #i4cs ※ @
21:54 @grammarware: The term I tried to remember from today’s morning: “high fidelity source code transformation” (the one that preserves whitespace & comment). ※ @
7:52 @grammarware: “Formal rules for comments are difficult enough to be easily forgotten to be included in a language standard” © #thesis ※ @
- 7:59 @GeePawHill: @grammarware aren't comments for precisely the stuff we can't express formally? if we could, they'd be called "code". #thesis ※ @
  - 12:34 @grammarware: @GeePawHill we still need to specify their structure, i.e. // or /*…*/ or %. Preservation of comments is desirable by some code refactorings ※ @
1:20 @grammarware: “Semi-automated means there can be significant improvement by using the right technology, which still must be operated by an expert” #thesis ※ @

5 Aug, 11:55 @grammarware: “The symbol "::=" is used to separate left from right.” © Donald Knuth. If I ever write a sentence like this, @notquiteabba will kill me. ※ @
- 8 Aug, 12:53 @notquiteabba: @grammarware Can't afford to kill any of my PhD students; there are just far too few and those who exist deserve to live! ※ @
  - 10 Aug, 17:52 @grammarware: @notquiteabba nice paradox: our scarcity is the only thing that saves us from extinction. ※ @

2 Aug, 21:54 @grammarware: A clever XSLT trick now prevents generation of broken links at uni-koblenz.de Seems like @cramforce was right twitter.com ※ @
- 23:31 @cramforce: @grammarware Just for the history books: I was trying to be ironic in the most cynic sense. XSLT is the biggest piece of fail technology evr ※ @
20:33 @grammarware: It should not be totally impossible to design a tool that takes an XSLT script and derives an XSD for its output. Far from trivial, though. ※ @

1 Aug, 14:04 @grammarware: During hacking I realised I need a new grammarware tool, called it “re-pretty-printer”. When something is already rendered & you change that ※ @
14:03 @grammarware: Pilot version of hypertext XBGF manual is online uni-koblenz.de A number of todos left, but generally sketched on top of existing stuff. ※ @

29 Jul, 17:25 @grammarware: Composing dictionary sections can be useful: I realised that what I used to call “grammar cropping” is what we call “grammar relaxation” now ※ @
12:37 @grammarware: Is there any kind of standard and/or researched pretty-printer for XML Schema? (Just another presentation of the same information). #ifi ※ @

9 Jul, 8:51 @grammarware: Funny benchmark: the amount of code a maintenance programmer has to look at in order to perform (seemingly) local re-engineering activities. ※ @
- 8:53 @GeePawHill: @grammarware closely related to xkcd's wtf-per-minute metric. ※ @
  - 8:55 @grammarware: @GeePawHill indeed! I guess it even correlates with a wtf-per-line-of-code metric. ※ @

7 Jul, 12:41 @grammarware: Model transformation = grammar transformation; model synchronisation = grammar convergence; model management = grammar engineering. #GTTSE ※ @

6 Jul, 12:43 @grammarware: Looks like in this age of SOA, MDE and meta-modeling people need to emphasize the importance of concrete syntax again and again. #GTTSE #IFI ※ @

25 Jun, 17:47 @grammarware: It is possible to have XBGF generators with side input: e.g., it could take a grammar AND, say, a codebase, to apply grammar inference. #ifi ※ @

24 Jun, 1:41 @notquiteabba: Finally done with my reviewing tasks for MODELS 2009. Tired. Disillusioned. Let me do some heavy Prolog hacking tomorrow. ※ @
- 13:30 @grammarware: @notquiteabba just curious, did you manage to find any use for my input for that? The third one was totally not of a kind I’d read otherwise ※ @
  - 14:08 @notquiteabba: @grammarware yeap, thanks Vadim for helping with MODELS reviews and some of the other conferences this year :-) ※ @

13 Jun, 1:33 @sternandy: Anyone know and good articles on DSL using functional programming? Why is F# better for creating these? ※ @
- 22:53 @grammarware: @sternandy Frankau, Spinellis, Nassuphis, Burgard, “Commercial Uses: Going Functional on Exotic Trades” lexifi.com #dsl ※ @
- 23:05 @grammarware: @sternandy also, ICFP 2009 accepted a language engineering paper by Newton and Ko describing a functional DSL, it might have some insights. ※ @
- 23:07 @grammarware: @sternandy finally, FP is freq. used in transformational DSLs. See e.g. JunGL progtools.comlab.ox.ac.uk XBGF uni-koblenz.de slps.sourceforge.net ※ @

12 Jun, 23:28 @grammarware: Just uploaded the final version of the extended abstract for #GTTSE about language convergence. #IFI And still not too late for a night out! ※ @

11 Jun, 18:50 @grammarware: XLDF is starting to produce bugs indicating some degree of inconsistency. Hacked through it for today, might decide to redesign later. #IFI. ※ @

9 Jun, 20:59 @furbach: Does any DSL-program finally can be translated into executable code? The example today at ADAPT was about configuration. #isweb #agki ※ @
- 21:04 @grammarware: @furbach grammar definition formalisms (SDF, ANTLR, even YACC) are DSLs that can be “compiled” to executable Java or C code. You meant that? ※ @
  - 21:27 @furbach: @grammarware But are there also DSLs which cannot be compiled? ※ @
    - 22:16 @grammarware: @furbach there are DSLs that are meant to be interpreted directly (e.g., Shell, PostScript, any advanced configuration language, arithmetic) ※ @
    - 22:17 @grammarware: @furbach …and there are DSLs that produce non-executable code, like modelling ones (the main reason some consider modelling languages apart) ※ @
      - 22:19 @radkat: @grammarware can you give me example of non-executable DSLs ? (mean, what they will be generated into) ※ @
        22:29 @grammarware: @radkat ECore is compiled to Java that has some abstract classes and interfaces plus “implementation” which does nothing—it was not modelled ※ @
        22:36 @radkat: @grammarware But nevertheless it's some framework, even frame*code*, that's intended to be used ? or else what for these artifacts are ? ※ @
        10 Jun, 0:13 @grammarware: @radkat indeed, miss. It’s just that vertical DSLs sometimes cannot express all the things that are needed to be executable at target level. ※ @
8:52 @grammarware: Another blog post about pains in Python. Matching operators ranting this time. spoonfulofhacking.blogspot.com ※ @

8 Jun, 15:07 @grammarware: RT @ashalynd: RT @bubbl_scala: Scala By Example : programming scala-lang.org — I wonder if “by example” books should be covered by LDF… ※ @

6 Jun, 12:48 @grammarware: A new blog post about binary operators and their imperfect implementation in Python: spoonfulofhacking.blogspot.com (and how to fix it!) ※ @
3:54 @grammarware: Quite happy with the schema for language documents:even the Gang’o’Four book (Design Patterns by @ErichGamma et al) fits here perfectly #IFI ※ @

5 Jun, 16:59 @grammarware: Meta-programming in a dynamic language such as python (meta-hacking?) is cool.Mad,but cool.One class did what I thought was a job for a DSL. ※ @
10:18 @grammarware: I’ll publish a series of small blog posts about the pains of #Python. The first one is about list methods: spoonfulofhacking.blogspot.com ※ @

4 Jun, 10:07 @grammarware: Published a very simple twitterparsing example: snipt.net — this one wins at the game of @twitbrain. #python ※ @
21:12 @grammarware: With my humble twitterparsing experiments I seem to inspire @notquiteabba to set up an aggregating account for our department: @institute4cs ※ @

2 Jun, 6:20 @grammarware: I woke up to our paper with @notquiteabba about Java grammar recovery and convergence being accepted for SCAM 2009. uni-koblenz.de Nice!! ※ @
- 7:48 @radkat: @grammarware Congratulations ! : ) ※ @
  - 21:35 @grammarware: @radkat thanks! You know we’ve worked hard for this ;) ※ @
    - 11:51 @radkat: @grammarware yeah, i know that this is a victory smelled with powder, so the sweeter it should be : ) ※ @
- 10:10 @corpaul: @grammarware congrats! ※ @

1 Jun, 22:49 @grammarware: TWIMC: all tweets from May 2009 that are related to yours truly are archived at uni-koblenz.de Parsers & generators is my hobby. Weird? ※ @
18:28 @grammarware: Another idea for a future XBGF generator that I missed before: find all top nonterminals in the grammar and reroot() to them. ※ @
- 4 Jun, 11:30 @grammarware: @grammarware implemented successfully yesterday, deployed to FL in SLPS slps.sourceforge.net Brought in nice ideas about convergence automation ※ @

31 May, 11:30 @grammarware: I like overview-ish papers when they are done well. dx.doi.org is a well-done one about XML compression. ※ @

28 May, 13:58 @grammarware: For completeness sake, we also need to converge the grammar extracted from #ECore generated from XSD with the manually created #ECore. ※ @

27 May, 23:58 @grammarware: Abstract types are tricky in ECore: why suddenly all these bits of inference wrt super types? ※ @
21:13 @grammarware: ECore2BGF extractor is finished (done in XSLT) and committed to SLPS slps.sourceforge.net As usual, I ended up implementing much more than min. ※ @
18:21 @grammarware: #Working on a grammar extractor from #ECore models… ※ @
- 18:23 @radkat: @grammarware jump to my SVN repositoir. : ) ※ @

19 May, 21:09 @grammarware: XBGF #todo: extract transformation that takes a selector name as a parameter. Given a:b x::c, extract(x) will give a:b x and x:c. XSD needs. ※ @

12 May, 13:22 @zef: Have to teach a class in 15 minutes. ※ @
- 13:41 @rickvanderarend: @zef what's gonna be the subject? ※ @
  - 16:41 @zef: @rickvanderarend it was about generating code from DSL for multiple software platforms (Java, .NET, PHP etc.) ※ @
    - 16:54 @rickvanderarend: @zef sounds interesting! ※ @
      - 17:02 @zef: @rickvanderarend tell me about it ;) ※ @
        18:04 @rickvanderarend: @zef the other way around seems more likely, does it not? :-) ※ @
        18:06 @zef: @rickvanderarend generating DSLs from multiple software platforms? ;) ※ @
        22:44 @grammarware: @zef Automated generation of DSLs, given the domain, has always been one of my dearest dreams. ※ @

10 May, 10:06 @grammarware: ' '.join(x.split()[1:]) vs x[x.index(' ')+1:] vs x.partition(' ')[2] #python #holywar #timtowtdi ※ @
- 10:09 @butaji: @grammarware а что если на regex переписать? ※ @

9 May, 10:38 @grammarware: I don't buy pure top-down design. Code-now-design-later is a good technique when you taste the domain before you think you're smarter in it. ※ @
- 11:01 @davengeo: @grammarware why design-code in some order first-last? Both options are unnatural. Code when you need, design when you need. ※ @
  - 11:30 @grammarware: @davengeo Agreed. However, one has to have a starting point somewhere, and coding/hacking seems like a perfect starting phase to me. ※ @
    - 11:36 @davengeo: @grammarware to start with a simple TDD "the first thing that works" is fine, then look at the design and refactoring... My way is no way. ※ @

8 May, 19:37 @grammarware: noticed that in all last tools I've made I always hack it up first until they work, then introduce OO where needed, optimise, add features. ※ @

5 May, 15:41 @grammarware: Будь у меня много свободного времени и мало дел, завёл бы твиттер для трансляции FMyLife/bash/… и умещал бы тамошние истории в 140 символов. ※ @
- 15:43 @ashalynd: @grammarware да ну, кому нужно это застеколье? :) ※ @
  - 15:59 @grammarware: @ashalynd боюсь, меня достаточно увлекал бы сам процесс усушки и утряски, чтобы о полезности не думать. Это я под впечатлением от @arjunbasu ※ @
    - 16:11 @ashalynd: @grammarware процесс был бы увлекательным, это да. но русский посложнее английского. это задача на кандидатскую :) ※ @
      - 16:21 @grammarware: @ashalynd на кандидатскую потянуло бы, если б у меня получилось этот процесс автоматизировать! ※ @
        16:22 @ashalynd: @grammarware да-да, и я о том же :) ※ @

28 Apr, 0:20 @grammarware: A good research engineer should always be solving problems in pairs or take only self-applicable problems to be able to validate his results ※ @

23 Apr, 2:31 @grammarware: if your transformation scripts are not suited for automated analysis, transform them automatically to transformation scripts that are. ※ @

21 Apr, 19:03 @grammarware: any grammar transformation with a precondition can be automated, a failed precondition would only mean inapplicability instead of failure. ※ @

14 Apr, 23:28 @grammarware: language specification is grammarware. ※ @

12 Apr, 7:27 @grammarware: is dropping all sections made from one included schema a valid automated transformation on a language document? ※ @

10 Apr, 1:19 @grammarware: another automation idea for grammar transformation/convergence: eliminate all unused nonterminals. ※ @

7 Apr, 20:16 @grammarware: language document evolution seems technically simpler than grammar evolution: rich transformation language allows to survive without DSL. ※ @
21:31 @grammarware: It's not «Algorithms + Data Structures = Programs» anymore, it's more «Data + Data Representation + Algorithms = Programs». ※ @
0:10 @grammarware: Since Java is such a mess anyway, why don't add some dynamic binding and type inference there? That'd be fun. ※ @

Generated by Twit2Html on 6 November 2009.