26

Re: call to arms

Hello, netch80, you wrote: N> When it is impossible to understand, something is function declaration or a variable, yet will not disassemble up to the end and will not read a context (the part from which is set already after this disputable line) is not "complexities with typification", it is parsing. And that here the difficult? Nitra such  without problems. Simply creates ambiguity and all. And further  understands. Though for LALR' it is finite the big problem. N> if you about possibility, for example, to collect in one rule some variants with priorities in between they are eliminated from similar possibilities - a certain general basic dial-up of possibilities undertakes. The minimum of possibilities between LL (1) and LALR is such poor polar fox that generally in any gate. More shortly transfer big  without a serious rewriting it is stupid a myth. N> who at you these general? To steam of names at least. GLR, GLL, Earley parser and some more then names I will not recall. N> In how to predict convenient extensibility of grammar without interpretation conflicts. 1) Parsim of ambiguity then we understand . In a syntax overload on types anything difficult is not present. From an overload of functions differs nothing. 2) Nitra implies creation of languages where syntax is library. Those the user can throw out out-of-date syntax just as now throws out out-of-date libraries. Moreover syntax can be switched on and off in one file. N> at the time of the Fortran could not think of by what it will be enveloped them "integer zz". I do not know that there in a Fortran happens. N> it is amusing. Coding style is not pleasant to me, but the idea is clear. It has been written for a floor of hour on a trick. N> anyway to an initial question about the book it is relations has no, well except for a long time already told - to add in the book three pages on TDOP would not worsen, but also did not refine. It would be possible as to add at level "on three pages" on each point, the book could increase easily in few times in thickness. They and so cut on live, it seems to me. Half of book devoted to parcers it was possible to throw out and replace stupidly here on these 3 pages. It is better than one lines TDOP than that that there is described.... <<RSDN@Home 1.0.0 alpha 5 rev. 0>>

27

Re: call to arms

28

Re: call to arms

Hello, netch80, you wrote: N> Well I, of course, understand the approach "after a parcer - though a flood because not our business", but there (around ), actually, and there is a main complexity and moreover, state that there the parcer and  is coroutines, and the parcer should enumerate logic of some sections. I am not surprised. If the parcer can build wood of trees it is not necessary to bind logic of a parcer and . Accordingly even with the registration of that that  it will be necessary to work with ambiguous nuclear heating plant it it will be easier than in a case with the bound logic. And the main complexities in  With ++ are connected not to ambiguities, and with that that type system in With ++ the bad. And its Turing completeness is not a principal problem. That with it to consult  it is necessary at once to start to write, how the interpreter of pure functional language with dynamic typification. N> Well here therefore I also say that the modern approach - to think over in advance how not to run into a pose that once it is necessary to pass on "general brakes" or manual . Good luck to you to invent as it to make on subset LL (1) and LALR. N> Not, out-of-date it is impossible to throw out. Here such task in 99 % of real cases. It is dogma of languages with monolithic syntax. In this case we add syntax at once all. Even that to whom it is not necessary. Also we take away it from all. Even that it is necessary to a clod. In a case with modular syntax we can add syntax even if it is necessary to one percent of users. Remaining it simply do not connect, and it will not touch them. Or to delete syntax from delivery by default. And those a floor of percent of users which still use it can return it reversely single line in a project configuration. N> so not in the Fortran, and after it. In a C with all clones. N> the description in the spirit of "int zz;" And later "int f (int, double);" leads to monsters of type "typedef typeof (typeof (int (*) (int)) (*) (int)) fn2;" (At the best, if it generally is compiled). N> And here  "type fp = function (int, function (double):double):int" no such problems has. N> here arguing by a row the Author: netch80 Date: 09.01 19:21 it well shows. If at us language on basis  that: 1) new syntax of types Is added. Nitra without problems can disassemble both variants simultaneously. 2) we write the code for an automatic rewriting of old syntax to the new. Besides on basis  it to make simply. 3) old syntax in the new is convertible. 4) old syntax is disconnected. This process can happen in flow of several years in different projects with different speed. Eventually there are shares of percent of reactionaries which refuse to pass to new syntax. Well and figs with them. Among themselves syntax generally does not influence integration of units in any way. N> as, will be though any ideas how to assume a similar rake at development of grammar and to avoid them? I do not make mischief at all, it is a subject which in any way how much I see, does not become covered neither the theory, nor the separated practice. Only experience of the developer of language. But if the error has been admitted, in a case with languages with modular syntax it can be corrected.... <<RSDN@Home 1.0.0 alpha 5 rev. 0>>