source: project/wiki/chicken-5-roadmap @ 32460

Last change on this file since 32460 was 32460, checked in by sjamaan, 5 years ago

Summary: Update roadmap: modular compiler is done

File size: 19.9 KB
1== CHICKEN 5 roadmap
5Here's a proposed list of things we would like to see in CHICKEN 5.
6Feel free to add more details if you know of a way to implement
7something or have an idea how to improve some part.  Please, no
8editing flamewars here!
10=== Modularising the compiler
12This work has been completed: the compiler now is composed of modules
13in the {{chicken-5}} branch (prefixed with {{chicken.compiler}}), but
14the following "nice to haves" are not yet implemented:
16* Define an "official API" for users of the compiler.  Basically everything that's currently being done through ugly {{##compiler#}} hacks should have a supported, documented way to do it.  Later, we can expose more features.
17** Hooks for adding new foreign types. Used by {{bind}}.
18** Hooks for adding new compiler literals?  Examples of this are the CHICKEN 4 {{numbers}} egg or when we want to turn srfi-4 into an egg.
19** Some standard way to determine the current source file (ideally this would be a library procedure which works the same way in compiled and evaluated code).  Used for things like the {{s48-modules}} egg.
20** Perhaps a way to define new compilation stages.
22=== Reworking the core modules ("units")
24Right now the modules supplied by core are somewhat arbitrarily named,
25and too many unrelated things are grouped together.  We should go
26through the system and look at what we have, then make logical names.
27Suggestion to appear later on this page, for further discussion.  We
28should attempt to align it with the r7rs naming conventions, to make
29things easy for that egg, and for people new to CHICKEN but familiar
30with other r7rs implementations.  This probably means "scheme" should
31be renamed and split up to "scheme.base", "scheme.load", etc.  A
32possible generalisation (or "convenience hack") could be to define the
33"scheme" module to import all of the underlying submodules.
35* As I've posted to the mailing list, I think using hyphen makes more sense than using dot.  --John Cowan
36** I think we're pretty much resolved to using dots, for various reasons appearing on the list (and because there's momentum in the other direction with e.g. the compiler modules). -- eh
38==== Replacing SRFI-14 with cset implementation from irregex?
40This has been discussed ages ago.  It might be more memory-friendly
41and performant.  One problem with the current SRFI-14 module is that
42it assumes Latin1 encoding (and therefore can only handle 256
43different characters), whereas most other CHICKEN components and eggs
44assume UTF-8.
46* Strong +1.  --John Cowan
47* Note that in the "Proposed removal from core" section below, srfi-14 is proposed to be removed from the core. --mario
50==== Refactoring the CHICKEN test suite to use a core library?
52As we remove a lot of cruft from core which it doesn't need, it may be
53a good idea to add some things that we ''do'' need.  Like the {{test}}
54egg: there is a lot of macro code duplication in core's test suite.
55It's probably better to ship a well-designed testing library with
56core, which core itself can also use.  This would make it easier, if
57we decide to do this later, to format test output on Salmonella in a
58consistent manner for both core and eggs.
60* That could even be done for CHICKEN 4, since it wouldn't break anything. -- mario
62==== Proposed libraries
64Let's follow R7RS for these:
66* chicken.base
68* chicken.char
69* chicken.complex
70* chicken.cxr
71* chicken.eval
72* chicken.file
73* chicken.inexact
74* chicken.lazy
75* chicken.load
76* chicken.process-context
78* chicken.repl
79* chicken.time (need this? want this?)
80* chicken.write
82What will we do with the SRFIs we implement?  It would make sense to
83define the following, but it would be tedious to import all these:
85; srfi-2 : and-let*
86; srfi-8 : receive
87; srfi-31 : rec
88; srfi-26 : cut, cute
89; srfi-17 : setter, getter-with-setter
90; srfi-10 : define-reader-ctor
91; srfi-39 : parameter objects
93* I'm planning to propose some of these (2, 8, 31, 26, 17) in a single R7RS-large library, probably called (scheme control) or (scheme control simple).  --John Cowan
94** Since this hasn't been standardised yet, and for improved compatibility and consistency with other Schemes, it's probably a good idea to define them as separate modules anyway.  Note that this does not preclude re-exporting them elsewhere as well. --Peter Bex
95*** I agree that they should be defined in their own modules and then reexported by some larger modules, e.g. some {{chicken}} library includes {{and-let*}}, etc. -- eh
97Also, is it {{srfi-2}} or {{srfi.2}}?  The latter would match up with
98{{(srfi 2)}} usage which is reserved by R7RS for SRFIs.
100* If we get rid of dots, then it's just {{srfi-2}} without special-casing it as the R7RS egg apparently does right now.
101** I don't quite understand this comment. {{srfi.2}} will map without special casing, is that what's meant by that? -- eh
103The list below is just a proposal, can be changed at any time.  We
104should also keep an eye on R7RS WG2, which may define a few things
105CHICKEN currently defines already.
107See also the concrete proposal taking shape
108[[/core-libraries-reorganization|here]], which subsumes the following.
110; chicken.modules : module, import, export, reexport, define-interface, module-environment, functor, use
111; chicken.types : {{:}}, the, assume, define-type, define-specialization, compiler-typecase
112; chicken.reader-extensions : set-read-syntax!, set-sharp-read-syntax!, set-parameterized-read-syntax!, copy-read-table, current-read-table (perhaps re-export define-reader-ctor?)
113; chicken.fixnum : fx+, fx-, fx/, fx*, fx<, fx<=, fx=, fx>, fx>=, fxand, fxeven?, fxior, fxmax, fxmin, fxmod, fxneg, fxnot, fxodd?, fxshl, fxshr, fxxor, fixnum-bits(?), fixnum-precision, most-positive-fixnum,  most-negative-fixnum, fixnum-bits, fixnum-precision, fixnum?
114; chicken.flonum : fp+, fp-, fp/, fp*, fp<, fp<=, fp=, fp>, fp>=, fpfloor, fpceiling, fptruncate, fpround, fpsin, fpcos, fptan, fpasin, fpacos, fpatan, fpatan2 (?), fplog, fpexp, fpexpt, fpsqrt, fpabs, fpinteger?, maximum-flonum, minimum-flonum, flonum-radix, flonum-epsilon, flonum-precision, flonum-decimal-precision, flonum-maximum-exponent, flonum-minimum-exponent, flonum-maximum-decimal-exponent, flonum-minimum-decimal-exponent, flonum?
115; chicken.syntax : er-macro-transformer, ir-macro-transformer, gensym(?), expand (is this useful at all?), get-line-number, strip-syntax.
116; chicken.bitwise : the subset of srfi-60 we support: bit-set?, bitwise-and, bitwise-not, bitwise-ior, bitwise-xor.  Possibly complete it with the remaining operations, and call it just "srfi-60"?
117; chicken.ports : The current stuff in ports, except for the string ports in scheme.base (also, see below).  Perhaps get rid of port-fold, copy-port, port, map, port-for-each?
118; chicken.exceptions (or srfi-12?  Would make more sense, but what about our extensions?  Put those in chicken.srfi-12?) : All the exception handling stuff.
119; chicken.load : If we want to keep them, load-noisily, load-relative, load-library
120; chicken.format : {{[fs]?printf}}, format (do we need this?), pp, pretty-print, pretty-print-width
122* If you put {{use}} in a module, how do you get access to that module?  I favor the R7RS solution, in which {{import}} does what Chicken {{use}} does, and is special-cased in terms of the module system so that it is always available.  --John Cowan
123** Right now, I think the module import/export forms are always available inside a module form.  This is no different from special-casing it, I think (unless I'm misunderstanding something). --Peter Bex
125* There will be R7RS (scheme fixnum) and (scheme flonum) modules.  I'm currently proposing to base the fixnums on R6RS and the flonums on {{math.h}} (not the egg of that name, but the whole C interface).  --John Cowan
126** That sounds like they'll be somewhat different from the list of identifiers we have.  And it will take a while before it's finalized I guess, so it's safer to define our own and later add the r7rs versions if we deem it acceptable. --Peter Bex
129==== Proposed removal from core
131The list below is just one hacker's idea of what could go.  Please add more.
133===== SRFIs
135SRFI-1, SRFI-13, SRFI-14, SRFI-18 might be removed. SRFI-69 will be
136removed, as discussed in CR #1142.
138As pointed out several times by John Cowan, SRFI-15 (fluid-let) is
139unsafe in the presence of threads, and any use is most likely broken
140and should be replaced with R7RS/SRFI-39 parameters.  Currently, core
141uses it in a few places, in a possibly dangerous way.
143Most importantly, there is no reason it has to be in core, because it
144uses only basic primitives.  I think it's best to delegate it to an
147===== queue datatype (data-structures), binary-search (data-structures), mmapped files (posix), object-evict (lolevel)
149Proposal already accepted in CR #1142.
151* I'm proposing a queue library for R7RS-large.  --John Cowan
152** It would be great if it could be inspired by CHICKEN's, but that's not strictly necessary, as there is plenty of room for multiple queue eggs --Peter Bex
154===== combinators
156Some of the combinators from data-structures are very nice, but there
157only a handful of them are actually useful.  There is no technical
158reason to keep them in core, they might fit better in an egg.
160* I'm proposing a similar library for R7RS-large.  --John Cowan
161** Maybe we can rip it out of core and wait for R7RS before implementing the egg. --Peter Bex
163===== Various ill-conceived POSIX things
165These things I don't like, but doesn't mean it *has* to go.  It may
166always be put in an egg of course.
168* file-select (but see the section about refactoring the scheduler!)
169* file-control (no need to be in core)
170* file-mkstemp (too tricky to use properly? maybe a different API)
171* file-read and file-write (too low-level)
172* file-stat (might be changed return a record type?)
173* set-file-position! (see the section on I/O refactoring)
174* All the time stuff.  It's too broken/difficult to use, and might be better off in an egg.  Core uses some of it, so we may need to reconsider and just improve the API.
175* terminal-name, terminal-port?, terminal-size (but chicken-status uses it!)
176* The process-stuff.  There are too many procedures which is confusing.  Boil it down to just one or two essential ones.  Possibly make a "fork&exec" implementation, which maps better to the Windows model, and still works fine on UNIX.
178===== Better API for continuations
180Nobody seems to use the "better API for continuations" by Feeley:
181continuation-graft, continuation-capture, continuation-return,
184If it doesn't benefit anyone (core doesn't use it, only two eggs do:
185shift-reset and continuations), it can be taken out.  It might be put
186into an egg.
188* +1 for an egg.  I'm going to propose this for R7RS-large.  --John Cowan
190* FWIW this seems to be pretty deeply-seated in core/runtime.c (to me at least!) -- eh
192=== Reworking the way libraries are loaded
194Right now there are just too many confusing things, like require, require-extension, use, import, load, load-library, require-library.
196* Import (with the function of use) should be the main API.  Load is necessary because it can load things whose names are determined at run time.  It should be able to load either source or binaries.   Include also belongs here.  --John Cowan
198Units and modules are confusing also.  This could just be a
199documentation issue.
201* Units should IMO be deprecated, with a compiler switch to turn off deprecation when compiling Chicken itself.  --John Cowan
202** I disagree: there's no reason why core should be "special" in any way.  We could de-emphasize their importance in the manual, instead. --Peter Bex
204* I disagree that units should be deprecated at all. I agree that import should be the primary API, with an alternative form for importing just identifiers (perhaps even {{(import-identifiers (foo bar))}}). -- eh
206==== Make the library load path a search path
208This keeps cropping up on IRC: people expect to be able to load libraries from their eggs using a search path containing multiple entries. This would allow you to {{(use ...)}} a module from your application without installing it as an egg.
210This is rather tricky: what happens when you compile it and install the whole program into some other location? Also, changing the way it's implemented is nontrivial, as it has been attempted before (see [[|#736]]).
212=== Refactoring the scheduler
214One missing ability in the scheduler is for threads to block on more
215than one object.  This would allow us to generalise {{file-select}} to
218=== Refactoring the I/O (ports) system
220Currently, ports are somewhat ill-defined: they're a hand-coded record
221type with a bunch of slots, with comments indicating which slot is
222used for what.  It would be cleaner and easier to understand the code
223if this was changed to a "proper" record type.
225The {{current-*-port}} identifiers should be rewritten to be proper
226parameters instead of fake ones which are rebound through fluid-let.
228Recently I discovered that set-file-position! does not work on string
229ports.  Port position should be part of the official interface, so
230that this is extensible, and if a port implements it, it can be
231rewound.  This makes sense at least for file-backed ports and string
234* Well, not all file-backed ports are seekable.  --John Cowan
235** That's okay; they can throw a "not implemented" exception. --Peter Bex
237This is also a good opportunity to look at why I/O is so slow.
239One small improvement I'd like to make is to change write-string to
240accept an offset into the string from which to write.  This would mean
241writing substrings does not have the overhead of first having to copy
242the substring to a new string and '''then''' writing it.  I ran into
243this once and I thought it was a shame, because it's such a trivial
244(but incompatible) modification.
246=== Integrating the full numeric tower
248This work has been completed: full support for the complete numeric
249tower is available in the {{chicken-5}} branch.  This includes support
250for literals in compiled code as well as full integration with the
253=== String encoding
255== Reject all NUL bytes
257If we reject all NUL bytes inside strings, we can encode strings more conveniently
258by adding a NUL terminator to all strings (nothing else changes).  If we do this,
259the FFI does not need to copy strings, which makes it much more lightweight.
261Things to look into:
263* What if the foreign code mutates the string and inserts a NUL?
264* How do we deal with the length?  Currently the internal operation is ##sys#size, which simply unmasks and returns the string's header.  The GC knows about this general principle.  By adding the NUL byte, we add another special case to the GC.  This is ugly and complicated.
265* Possibly the operations we support on blobs need to be extended, so that all current abuse cases for strings can be handled by blobs.
267== Unicode
269This at least needs some additional thought.  Do we want to make UTF-8
270the "official" encoding?  If so, ideally, all string operations should
271reject invalidly encoded byte sequences (should we still allow NUL
272bytes to be represented?).  What to do with the Unicode case folding
273lookup tables, string-ref?
275* Go full Unicode.  If Chibi can do it, so can we.  R7RS is factored to push the big Unicode tables into (scheme char).  However, IMO the NUL character is completely worthless as a character: it has no semantics worth mentioning.  We can forbid it in strings, as R7RS-small allows.  --John Cowan
276** Seems sensible. --Peter Bex
278If we go full Unicode, the SRFI-4/blob types might need some
279attention, because strings can no longer be (ab)used as byte vectors.
281* Why are there both u8vectors and blobs?  IMO they should be the same thing, and should be R7RS bytevectors.  I'm working on a R7RS-large numeric vector library that allows either SRFI 4 style (separate data types for different kinds) or the style used in later SRFIs and R6RS (everything is just a view on top of bytevectors).  --John Cowan
282** u8vectors are less "core" than blobs (which is a consequence of the low-level representation).  In fact, we might be able to take srfi-4 out of core. --Peter Bex
284=== Improve the egg system
286Since this is a rather comprehensive point, there is now a
287[[chicken-5-roadmap-egg-system|separate document for it]].
289=== Make set!'ing of unbound variables an error
291R7RS recommends making this an error for modules but allowing it in the REPL.
293* We already check for renaming already bound identifiers, maybe that's not so hard after all. I will investigate this --Christian Kellermann
295=== Determine how to make CHICKEN 4 eggs live alongside CHICKEN 5 eggs
297Currently, "THE SYSTEM" does not have any special considerations for
298the major CHICKEN release used.  This could be considered an
299oversight.  To make it possible to continue using CHICKEN 4 eggs while
300CHICKEN 5 is being developed and matured, there needs to be some sort
301of way to do this.
303Currently, we have the master list of available eggs, which lives in
304the svn repo.  THE SYSTEM is extremely simple and doesn't really care
305much about how eggs are supplied, so we could just fire up a second
306instance of henrietta-cache which fetches from a ''different'' master
307list containing the CHICKEN 5 eggs.  However, what can we do to make
308life easier for egg maintainers?
310The official CHICKEN egg repo (SVN) already has taken care of this due
311to the {{/release/N}} namespacing.  The thing that needs to be changed
312is the location of the henrietta CGI, to include a version number, or
313we could add an extra URL parameter and teach it about the versions.
315For user repos, a simple way is to simply start a second repository
316and call it a day.  However, this will probably result in awkward
317names.  Making a new branch results in the same problem: the master
318branch would correspond to an outdated release!
320==== The simplest approach: just carry on
322Just continuing in the old repository is possible, if no new releases
323need to be tagged for the old CHICKEN release.  This mostly precludes
324emergency bugfix releases, but these could be continued on a different
325branch (release-info only takes into account tarballs which get
326generated from a tag name, after all!).
328To prevent version tag clashes, the egg's major version should be
329bumped for CHICKEN 5.  Let's take for example an egg which has
330released 1.0, 1.2 and 1.3 for CHICKEN 4.  If we bump the major
331version, we can release 2.0, 2.1, etc for CHICKEN 5.  If an important
332bugfix needs to be made for the CHICKEN 4 version, we can continue
333with 1.4.  If we don't bump the major version, the egg would be forced
334to use micro version numbers for those, like 1.3.1.  Both approaches
335are fine, depending on how much effort is expected to be put into the
336"old" branch.
338The old release-info file will be untouched and continue to be used by
339the CHICKEN 4 version of Henrietta-cache.  For CHICKEN 5, a new file
340is made (ie myegg.chicken-5.release-info) which starts out empty, and
341as new releases are made will continue with the number where the
342CHICKEN 4 branch left off.
344==== Rework each egg's release namespace
346Another, possibly cleaner, approach is the following:
348* When an egg is ported to CHICKEN 5, rename or copy all existing tags,
349prefixing them like {{chicken-4/1.2}}, for example.
350* Make a chicken-4 branch from master and update the release-info
351file's location in the master egg list for CHICKEN 4.
352* Clear the release-info file in master, and submit its location for
353inclusion in the master egg list for CHICKEN 5.
355This way, new eggs and old eggs will ''always'' have the master branch
356point to the active version.  It does mean a little bit more work on
357every major release.
359To avoid having to clear the release-info file every time, we could
360also extend it to include a major release version number (and if it's
361missing, assume "4"?).  This means the release-info file would list
362both CHICKEN 4 and CHICKEN 5 (and later CHICKEN 6) releases in the
363same file.  This might make maintenance a little easier, but requires
364a small change in henrietta-cache.
366* IMO the brains should be in the henrietta web API.  --John Cowan
367** I don't think that's necessary.  In any case, there '''must''' be some way for the egg authors to indicate for which CHICKEN version the egg is. --Peter Bex
Note: See TracBrowser for help on using the repository browser.