wiki:WG1BallotCowan

Version 19 (modified by cowan, 6 years ago) (diff)

--

Instructions

  • You may list as many of the options as you want in order of preference.
  • Options are comma-delimited (ignoring space) and case-insensitive.
  • You can pipe-delimit (|) options you want to give equal weight to.
  • You may write in your own option if you announce it to the list first.
  • You may specify a variant with option/variant, for example srfi-1/module to vote for srfi-1 but clarify it should be in a separate module. Please also include the srfi-1 option in this case.
  • You can write a free-form rationale after the "preferences" line,
  • module means "yes, but I want it in a separate module",
  • wg2 means "no, but I think it should go in WG2".
  • undecided means I want to discuss this issue further.
  • Abstain on any item by leaving the preferences blank.

WG1 Ballot Items To Finalize by July 1st

Previous Undecided and Re-opened Ballot Items

#32 user-defined types

Do we support any means of creating disjoint user-defined types, such as in SRFI-9, SRFI-99 or the R6RS record system?

WG1 voted srfi-9 before. New arguments against filter constructors were raised, so the ticket was re-opened.

References:

Rationale: The most important thing about SRFI 9 is how pervasive it is (26 out of 30 implementations I track at http://tinyurl.com/scheme-s5 support it). Yes, it's messy. Yes, we should keep it.

#28 binary I/O ports

Do we provide any binary input or output ports, and if so how do we construct them and operate on them? Can binary and textual operations be mixed on the different port types?

PortsCowan provides binary port operations along with other extensions.

R6RS provides an entirely new I/O system, as well as a separate R5RS-compatible I/O system.

The withdrawn SRFI-91 provides yet another I/O system supporting binary ports.

Note this item as well as #29 and #31 specify semi-orthogonal aspects of I/O systems which are typically specified together by individual proposals. If the same proposal doesn't win for all three, the aspects will be merged as needed.

WG1 voted weakly in favor of cowan before.

#83 Auxiliary Keywords

In R6RS auxiliary keywords (such as else in cond and case forms) are explicitly exported from the (rnrs base (6)) library. Do we want to bind and export these from the core library?

If else is bound in the default module, then it must be imported at the call site whenever using it in cond or it won't match hygienically.

If else is not bound in the default module, then it must not be bound or imported at the call site whenever using it in cond or it won't match hygienically.

Another option is to specify for cond and case that they match the else identifier literally, ignoring any hygiene. This breaks compatibility with R5RS and R6RS.

WG1 voted unbound previously. New issues were brought up on the list so the ticket was re-opened.

References:

  • Options: bound, unbound, unhygienic, undecided
  • Default: unbound
  • Preferences: bound

Rationale: The arguments for binding this convince me.

#3 module naming convention

We need a naming convention for the core modules and standard libraries of the new module system.

The existing break down is based on John Cowan's earlier proposal of factorings in items #71, #72, #73, #74, #75, #76, #77, as well as an I/O module breakdown in PortsCowan. There have been various tickets proposing changing this, so we are re-opening the ticket.

Note: The (scheme io) module in the draft was never voted on and doesn't belong there.

New Ballot Items

WG1 - Core

#85 Blobs, bytevectors, byte-vectors, octet-vectors, or something else?

Now that we have blobs, we have to decide what to call them. R6RS uses bytevector, SRFI-4 and SRFI-68 uses u8vector, while the original WG1 proposal used blob (which is therefore the default).

  • Options: blob, bytevector, byte-vector, u8vector, octet-vector, undecided
  • Default: blob
  • Preferences: blob

Rationale: I call them blobs because I see them as objects whose size is measured in bytes, not specialized vectors of 8-bit unsigned integers. It's easy to layer such an interpretation, along with many other interpretations, on top of blobs, but blobs should themselves remain simple for WG1.

#118 Simple literals must be explicitly delimited.

In R5RS syntax such as #t#f is left unspecified - some readers may parse this as the true literal followed by false. R6RS requires identifiers, characters, booleans, number objects, and . to be terminated with a "delimiter" or by the end of input.

References:

  • Options: delimited, unspecified, undecided
  • Default: unspecified
  • Preferences: delimited

Rationale: Allowing #true to mean #t rue is just silly. If implementations want to accept #true, they should be able to do so.

#119 Whether to treat # as a delimiter.

In R5RS foo#f is a valid identifier, whereas R6RS requires # to act as a delimiter, so that this would parse as the identifier foo followed by the false literal.

  • Options: delimiter, no, undecided
  • Default: no
  • Preferences: no

Rationale: Backward compatibility rules here. Chicken depends heavily on identifiers with embedded #s.

#123 Extend unquote and unquote-splicing to multiple arguments

This is a change also made by R6RS (and CL).

References:

  • Options: multiple, single, undecided
  • Default: single
  • Preferences: single

Rationale: I'm not convinced this is useful enough.

#124 Nested quasiquote semantics

References:

  • Options: r6rs, r5rs, undecided
  • Default: r5rs
  • Preferences: r6rs

Rationale: R6RS is better defined, and we should accept it. The vagueness in R5RS helps nobody.

#125 Allow procedures not to be locations (making EQV? unspecified in some additional cases)

This is a change also made by R6RS.

  • Options: r6rs, r5rs, undecided
  • Default: r5rs
  • Preferences: r6rs

Rationale: Treating procedures as locations distinguishable by eq? is a hack nobody ought to depend on.

#126 Partly specify the mutability of the values of quasiquote structures

This is a change also made by R6RS.

  • Options: r6rs, r5rs, undecided
  • Default: r5rs
  • Preferences: r6rs

Rationale: This is a useful provision and doesn't constrain anyone much.

#127 Specify the dynamic environment of the before and after procedures of dynamic-wind

R5RS is slightly ambiguous, saying

BEFORE is called whenever execution enters the dynamic extent of the call to THUNK and AFTER is called whenever it exits that dynamic extent.

without saying clearly whether before and after themselves are called before or after the dynamic extent is entered or exited.

  • Options: outside, inside, unspecified, undecided
    • outside: called outside the dynamic extent (R6RS)
    • inside: called inside the dynamic extent
    • unspecified: R5RS
  • Default: unspecified
  • Preferences: outside

Rationale: I think outside was the intention of the R5RS authors and this should be clarified.

#135 let-values and let*-values

These R6RS procedures were part of #77 (modularization of multiple values), but were never explicitly voted up or down by WG1, so I'm opening a new ticket for them.

  • Options: yes, no, undecided
  • Default: no
  • Preferences: yes

Rationale: Some implementations can optimize these over calls to call-with-values, and they are much more convenient most of the time.

#137 Current-seconds semantics still open

In issue #70, WG1 voted to make current-seconds an optional procedure, but there is no guidance about what it returns. The second choice, TimeCowan, calls it current-posix-second and makes it return the integral number of seconds since the Posix epoch, excluding all leap seconds. We need a final decision on name and semantics.

Need proposals.

  • Options: cowan
  • Default:
  • Preferences: cowan

Rationale: That is, accept the TimeCowan view as noted above.

#147 Allow literal file spec lists in include and include-ci

This will be necessary on implementations where some files don't have character-string names.

  • Options: yes, no, undecided
  • Default:
  • Preferences: yes

Rationale: The issue here is that not all valid filenames are valid character strings: in Posix, filenames are sequences of 8-bit units excluding 0x00 and 0x2F, and on Windows they are sequences of 16-bit units with more exclusions. Those don't necessarily translate one-to-one UTF-8 and UTF-16 respectively, or any other particular encoding. The language of file specs used by open-*-file and friends allow implementation-specific extensions to cover these problems; include(-ci) should allow the same, as should load.

#148 Allow include-ci at top level

Currently include-ci is allowed in a module but not at top level, as include is. That's an arbitrary restriction that should be removed.

  • Options: yes, no, undecided
  • Default: no
  • Preferences: yes

Rationale: That is, allow it and remove the restriction. The point of include-ci is to bring in files of R5RS code that uses case typographically rather than orthographically.

#149 blob ports

We've voted to add string ports, which are character ports backed by Scheme strings. Since we have blobs another potential extension is blob ports, which binary ports backed by blobs. These are described in PortsCowan, but it's unclear if they were specifically voted for or against in the previous ballot.

  • Options: cowan, none, undecided
  • Default: none
  • Preferences: cowan

Rationale: Blob ports allow the construction and deconstruction of blobs in the same way that string ports allow the construction and deconstruction of strings, and should be provided for the same reasons.

#150 cond-expand at top level

Currently cond-expand is only valid in a module. Should we allow it at top level in a program?

  • Options: yes, no, undecided
  • Default: no
  • Preferences: yes

Rationale: Scripts need the ability to decided which modules to load (or which code to include) based on either the presence of other modules or specific criteria about the implementation, just as modules do.

#153 Renaming blob procedures

The blob procedures don't follow the same system as the rest. I propose these changes:

copy-blob => blob-copy
copy-blob! => blob-copy!
partial-blob => blob-copy-partial
copy-partial-blob! -> blob-copy-partial!
  • Options: new, original, remove, undecided
  • Default: original
  • Preferences: new

Rationale: The existence of the old names just shows that I didn't consider blobs and CompleteSequenceCowan in the same breath. Let's fix that while we can.

#154 Physical newline in a string equivalent to \n (that is, U+000A)

R5RS leaves this situation undefined, but R6RS, CL, and most languages that allow it (C does not) treat physical newline and escaped newline as equivalent, even if the local representation of line endings is \r\n or U+0085 or what not.

  • Options: unix, local, unspecified, undecided
  • Default: unspecified
  • Preferences: unix

Note: There is nothing Unix about this except that Unixen have the same convention. As noted, many other languages prescribe it.

#155 Make recursively defined code an explicit error

Allowing examples like these will make code-walkers (including compilers and interpreters) excessively complicated:

#1=(begin (display #\x) . #1#)

(lambda #2=(a b c #2#) ...)

(+ . #3=(1 2 3 . #3#))

  • Options: error, unspecified, undecided
  • Default: unspecified
  • Preferences: error

Rationale: This does not mean that the error must be signalled (raised). It just means that Scheme programmer shouldn't be allowed to depend on these tricks working, for some value of "working".

#156 Replace "an error is signalled" with "an implementation-dependent object is raised as if by raise"

The following situations are described in draft 1 (and R5RS) with "an error is signalled":

  1. The file-spec given to call-with-input-file, call-with-output-file, open-input-file, or open-output-file represents a file that cannot be opened.
  1. An end of file is read from a port by read after the beginning of an object's external representation, but the external representation is incomplete and therefore not parsable.

Rationale: I propose that in both cases the implementation be required to raise an exception as if by applying raise (that is, non-continuably) to an implementation-defined object, which means it can be caught by the R7RS exception system. Note that there is no requirement to create a fresh object.

  • Options: signal, unspecified, undecided
  • Default: unspecified
  • Preferences: signal

Rationale: I am not in favor of forcing Schemes to signal exceptions where they were not required to do so in R5RS. But in those few cases where implementations already must signal an exception, they should signal it so that it can be caught using the standard exception catcher. The overhead of doing so should be nearly nil.

#162 Remove DELAY and FORCE altogether

They are present in R4RS and R5RS, but not IEEE Scheme (which is our baseline). There are problems with a straightforward implementation that SRFI 45 fixes, but we voted down SRFI 45. Given that, we should consider removing them from the standard altogether. (Of course this does not mean compliant implementations can't provide them, it just means they won't be in a standard module.)

  • Options: remove, keep, undecided
  • Default: keep
  • Preferences: lazy, remove

Rationale: Alex added lazy (i.e. SRFI 45) back to the ballot, so I voted for it. Failing that, I really don't want to standardize something that's known to be broken in existing non-SRFI-45 implementations. They don't really provide lazy Scheme, and those who want actual Lazy Scheme know where to get it.

#164 Meaning of char-numeric?

The current draft, like R6RS, defines char-numeric? according to the nonexistent Unicode Numeric property. That has to be fixed. Options:

  1. char-numeric? returns #t if the character's Numeric_Type property value is other than None. This means that many hanzi are both alphabetic and numeric.
  1. (Omitted, because it does not preserve IEEE Scheme)
  1. Define char-numeric? to return #t only for 0, 1, 2, 3, 4, 5, 6, 7, 8, and 9. This retains compatibility witht R5RS, and we can still use char-numeric? to parse numbers, and safely use `(- (char->integer c) (char->integer #\0))` to obtain the digit value the character represents. (Note: R5RS programs that use char-numeric? to parse numbers will break if we adopt the current draft's definition of char-numeric?). Gauche, Gambit, and Chicken (without the utf8 egg) work like this.
  1. Define char-numeric? as equivalent to the Numeric_Digit property (general category value of Nd). Guile 2.0, Kawa, Larceny, Ypsilon, Mosh, and IronScheme? work like this.
  1. define char-numeric? as equivalent to the Number property (general category values of Nd, Nl, No). Scheme48, Chez, and Ikarus work like this.
  • Options: any, number, digit, ascii, undecided
  • Default: ascii
  • Preferences: digit

Rationale: There is no advantage to allowing U+0660 (Arabic-style 0) to be called non-numeric. Of course, there is no requirement for an application to support this character, but if it does support it, it should tell the truth about it, just as Arabic letters should be reported as alphabetic. On the other hand, letter-based numbers, fractions, Roman numerals don't really fit the traditional profile of numeric characters, which is why the expansive number and any choices don't seem appropriate.

#166 Add predicate and accessors for error objects

(Email from Vincent Manis)

Problem: It's impossible to write a portable error handler that writes out the message and irritants that were passed to error.

This comes about because error creates an "implementation-defined object". I would assume that this hides the whole exception class hierarchy a WG2 implementation might provide. Since the message and irritants arguments to error are presumably living in this implementation-defined object, it should be simple enough to provide accessors to extract them, so that the above "portable error handler" can be written.

Suggestion: Add the following procedures:

(error-object? object)

Returns #t if object is something created by error, #f otherwise. Any constraints on type disjointness are up to the implementation.

(error-object-message object)

Returns the message of object.

(error-object-irritants object)

Returns a list of the irritants of object.

  • Options: manis, none, undecided
  • Default: none
  • Preferences: manis

Rationale: I understand Vincent to mean that error-object? returns #t if the object could have been created by error, not that it necessarily was. So what is guaranteed is that objects raised by error satisfy error-object?, not that nothing else satisfies it.

#167 Add constructor for error objects

(Email from Vincent Manis)

Problem: Raising arbitrary objects as exceptions has been found to be nasty in some other languages (Python and C++ in particular).

This one is a tad speculative, but I'm reluctant to encourage people to write things like (raise 4), because of course it doesn't respect any module boundaries. I think the intent with the descriptions of raise and raise-continuable was to allow exception hierarchies to be added in WG2 without constraining them here. I would suggest adding a new procedure:

(make-error-object message obj ...)

to creates the implementation-defined object error is supposed to create, and adding a sentence to the raise and raise-continuable entries that says "The effect of applying this procedure to an object not created via make-error-object is implementation-defined." This allows WG2 to do what it wants regarding exception objects, and to limit the types of exception objects allowed, without breaking anything in WG1. Error can be defined as:

 (define (error message . objs)
   (raise (apply make-error-object message objs)))
  • Options: manis, none, undecided
  • Default: none
  • Preferences: none

Rationale: I agree that developers should document what sort of condition objects their APIs raise, and that raising numbers is a Bad Thing. But the fact is that existing code and implementations raise all sorts of things as condition objects: in SRFI 12 (which Chicken uses), they are lists. Not everything can be stuffed into the message-plus-irritants paradigm.

#169 Add standard-*-port procedures

These return the initial values of the corresponding current-*-port procedures, and can be used to access the implementation-provided standard input, output, and error streams.

  • Options: r6rs, none, undecided
  • Default: none
  • Preferences: none

Rationale: As Alex says, these are tricky to get right, and anyway the description here does not match R6RS, which requires the standard-*-port procedures to return fresh ports. Let's not go there.

#171 Duplicate identifiers in define-record-type

What happens if define-record-type is specified with two fields that have the same accessor identifiers provided for both fields? More generally, we need to say what happens when any two identifiers are non-unique.

This ticket deals specifically with the situation where two identifiers (accessors or mutators) of two field clauses in a define-record-type form are identical. This is not meant to address field names and what happens or what it means if the field names are symbolically equivalent but lexically distinct.

  • Options: error, unspecified, undecided
  • Default: unspecified
  • Preferences: error

Rationale: This makes no sense; it ought to be an error (that is, no conformant Scheme program should rely on what happens in such a case).

#173 Unifying BEGINs

In R5RS, there are three kinds of BEGINs:

1) All subforms are expressions; this can be used wherever an expression can be used. (4.2.3)

2) All subforms are definitions; this can be used wherever an internal definition can be used. (5.2.2)

3) Subforms can be definitions or expressions intermixed in any order; this can be used only at top level. (In R7RS we extend this to module top level as well). (5.1)

In particular,

(define (x)
 (define y 32)
 (begin
   (define z 45)
   (set! y z))
 y)

is not licensed by any of these provisions, and consequently is not valid R5RS Scheme. Nevertheless, all of my usual Schemes accept the above definition except Scheme48/scsh and SSCM -- actually, SSCM fails when you invoke x rather than when you define it. So I'm proposing that we unify them for R7RS.

  • Options: cowan, r5rs, undecided
  • Default: r5rs
  • Preferences: r5rs

Rationale: I'm no longer sure I believe my own reasoning here. Since such a begin can only be used at the boundary between definitions and expressions, there's little sense in a macro returning it, as such a macro could only be used there either.

#174 Safe uses of multiple values

Currently, uses of values where the values are discarded anyway is illegal, but all the usual Schemes except SCM and SSCM accept them (I tested with begin). Should we go with something close to the R6RS wording?

"The continuations of all non-final expressions within a sequence of expressions, such as in lambda, begin, let, let*, letrec, letrec*, case, and cond forms, take an arbitrary number of values."

The definition of begin would need to change too:

(define-syntax begin
  (syntax-rules ()
    ((begin exp)
     exp)
    ((begin exp1 exp2 ...)
     (call-with-values
         (lambda () exp1)
       (lambda args
         (begin exp2 ...))))))
  • Options: safe-values, r5rs, undecided
  • Default: r5rs
  • Preferences: safe-values

Rationale: What makes obvious sense, and what almost all implementations permit anyway, should be permitted by the standard. That is, programmers should be able to rely with confidence on its working, rather than just hoping they don't stumble across an implementation where it happens not to work.

#45 Record-let syntax and semantics

(record-let <record-data> ((<variable> <field>) ...)
  <body>)

Where each <variable> is filled with the corresponding data <field> from <record-data> as in a <let> expression, then the <body> is evaluated with these bindinds added and last expressions is returned. It is an error if the <record-data> does not contain corresponding <fields>.

Notice that this works directly on the data itself and that the data may contain more fields than the one cited in the record-let expression allowing code to be reused for inherited records.

Do we need to be able to check at runtime if a given record data has a given field ?

  • Options: record-let, none, undecided
  • Default: none
  • Preferences: none

Rationale: This is actually very nice; in particular, it is hygienic, unlike with statements in many languages.. However, it is primitive syntax (syntax-rules cannot implement it unless it has low-level access to the fields of a record) and it is a novelty. Let's push this to WG2.

#172 Multiple returns from map

R6RS specifies that map does not mutate previous results if there are multiple returns from map. Should we include this language?

  • Options: r6rs, unspecified, undecided
  • Default: unspecified
  • Preferences: r6rs

Rationale: This prevents an implementation based on directly mutating a string, vector, or list, but I think the results are better.

#178 Shadowing with internal definitions

From Andre Von Tonder:

On p 19, some shadowing problems that would break lexical scope are declared to be errors. However, I believe there are other examples that shold be errors that are not covered by the report. In R6RS a more general criterion was used - please see R6RS for details. Here is an example that does not violate the WG1 report but should be an error becasue it violates lexical scoping. It does not violate the WG1 criterion because the meaning of x is not needed to determine whether (foo x p ) is a definition.

    (let ((x #f))
      (let-syntax ((foo (syntax-rules (x)
                          ((_ x y) (define y 'outer))
                          ((_ _ y) (define y 'inner)))))
        (let ()
          (foo x p)
          (define x #f) ;; this should be an error because
                        ;; it shadows the previous line where
                        ;; x has already been used in its outer sense
                        ;; during expansion
          p)))

Here is another example that WG1 allows but that would cause violation of lexical scoping, because the macro would be evaluated first and treat ... as a placeholder in a region where it is shadowed to be the variable bound to 1:

    (let ()
      (define-syntax list-macro
        (syntax-rules ()
          ((_ x ...) (list x ...))))
      (define ... 1)    ;; This shadows ... in previously expanded macro
                        ;; body and will be a violation of lexical scoping
      (list-macro 1 2)) ;; if the last line evaluates to (1 2)

OTOH, it is unclear to me if WG1 allows this or not.

    (let ((x #f))
      (let-syntax ((foo (syntax-rules (x)
                          ((_ x y) (define y 'outer))
                          ((_ _ y) (define y 'inner)))))
        (let ()
          (define x #f)
          (foo x p)
          p)))
  • Options: r6rs, r5rs, undecided
  • Default: r5rs
  • Preferences: r6rs

Rationale: I'm more willing to believe that the R6RS people got this right than that we did.

WG1 - Modules

#112 REPL redefinitions

R5RS leaves unspecified the semantics of redefining a standard binding in the REPL. Do we want to specify semantics, or some set of allowed behavior for this in the WG1 standard?

REPLs may allow redefinition. The four cases that occur are redefining to/from syntax/non-syntax locally/imported, and the issue is what happens to previous references to the definition. The general possibilities are:

  1. redefinition signals an error
  2. previous references are overridden (generally not possible if it the previous definition was syntax)
  3. previous references are preserved (indicating a new binding was created, often preferred if replacing non-syntax with syntax to avoid runtime errors)
  4. the semantics are left unspecified

So all 64 combinations for these 4 values in the following 4x4 matrix are feasible:

From/To? import import syntax define define-syntax
import ? ? ? ?
import syntax ? ? ? ?
define ? ? ? ?
define-syntax ? ? ? ?

Not all 64 combinations necessarily make sense. The default from R5RS is "unspecified", which means all 16 values are unspecified.

  • Options: override, preserve, common, unspecified, undecided
    • override: override for all 16 values (non-syntax to syntax can break closure references)
    • preserve: preserve for all 16 values (must always create a new definition, not mutate, contrary to most implementations)
    • common: most common behavior among implementations - override, except preserve for non-syntax to syntax
  • Default: unspecified
  • Preferences: common

Rationale: The common behavior is provided by all implementaions except SCM and Wraith. We should standardize it.

#132 Imports override previous imports?

The current draft describes importing different bindings for the same identifier as "an error." R6RS explicitly requires this to signal an error. Do we want to change this?

This ticket refers only to modules - the top-level semantics are decided in ticket #112.

  • Options: override, preserve, error, unspecified, undecided
  • Default: error
  • Preferences: error

Rationale: This means that to override you have to declare the things overridden, which is probably good. Silent overriding is very ugly, as is silent non-overriding. High-quality implementations should make this a static error.

#160 Interleaving of imports and code in a module

Given

   (module (name)
     (begin c1 ...)
     (import (A))
     (begin c2 ...)
     (import (B))
     (begin c3 ...))

the intention, reference implementation, and specification from Scheme48 on which the syntax was based say that all imports establish the initial environment and then the code is expanded in order, but interleaving the imports is conceivable.

  • Options: shinn, interleave, unspecified, undecided
  • Default: shinn
  • Preferences: shinn

Rationale: Conceivable but not worthwhile.

#163 Allow modules at the REPL?

Should users be allowed to enter a module form at the REPL?

Note that there are actually many varying approaches to generating moduls at runtime, and Scheme48 and Chibi use an out-of-band REPL operation to create new modules, leaving the module binding open.

  • Options: yes, no, undecided
  • Default: no
  • Preferences: no

Rationale: Given the static nature of our modules, I can't see doing this. Modules are for programming in the medium to large scale, or for making libraries.

#141 What are the semantics of modules with respect to separate compilation?

ModulesShinn says that the bodies of libraries are evaluated before any of the bodies of the importing library; does that include, eg, "at compile time" rather than at "run time"? It's not clear.

  • Options: compile-time, unspecified, undecided
  • Default: undecided
  • Preferences: unspecified

Rationale: What happens at compile-time stays at compile-time.

#158 mutating imports

Currently the semantics of calling set! or define on an imported binding is undefined. Do we want to specifically make this an error?

  • Options: error, allowed, unspecified, undecided
  • Default: unspecified
  • Preferences: error

Rationale: Yes, for the same reasons as #132.

#159 base environments

What is the base environment provided by the repl, scripts, and the result of (scheme-report-environment 7)?

The intention was the base script environment was empty, scheme-report-environment was (scheme base), and repls were an implementation-defined superset thereof, but there are other options and we need to clarify this.

  • Options: shinn, undecided
    • shinn: intention as described above
  • Default: shinn
  • Preferences: shinn

Rationale: I don't see anything better than this.

#161 module argument to eval

It would be useful to allow modules as an argument to eval in addition to environments. This could be done with a special syntax, or just the module name as a list.

R6RS provides a procedure environment which just takes a list that looks like an import spec an generates the corresponding environment.

  • Options: r6rs, none, undecided
  • Default: r6rs
  • Preferences: r6rs

Rationale: Actually, it takes any number of such arguments. This allows eval to do its work against anything specifiable as a module, whether it includes the base or not.

#139 exit

The ballot statement for #62 said we had voted for exit when we voted for ModulesShinn, but that page doesn't mention exit. So we need to vote on it.

  • Options: yes, no, undecided
  • Default: yes
  • Preferences: yes

Rationale: Yes. Every implementation provides it.

#144 strip prefix on import

I'm thinking that for importing code that defines its external symbols as foo:this, foo:that, and foo:tother, there should be a type of import clause that strips a specified prefix from imported symbols. This is equivalent to renaming on import or renaming on export, but less painful, in the same way as the prefix import clause does.

Specific proposal: (strip-prefix <import-set> <prefix-identifier>).

  • Options: yes, no, undecided
  • Default: no
  • Preferences: yes

Rationale: This is especially nice if you are importing numeric libraries: you can use + - * / with non-standard definitions, but if straightforwardly imported they will not collide with standard + - * /.

WG1 - I/O

#133 Provide read-line

This is an R6RS procedure that was part of PortsCowan, but never explicitly voted up or down by WG1. It reads a single line up to a line delimiter from a given port (the current input by default) and discards the line delimiter.

  • Options: yes, no, undecided
  • Default: no
  • Preferences: yes

Rationale: Hand-rolled implementations will tend to handle only the line delimiter used by default in the programmer's environment, which does not generalize. In particular, it's now common to have text files with different types of line endings in the same environment.

#170 Add with-error-to-file procedure

Since we now have current-error-port, arguably we should have with-error-to-file for completeness.

  • Options: yes, no, undecided
  • Default: no
  • Preferences: no

Rationale: I see no point in making it easy to redirect the standard error.

#176 Are string ports exclusively character ports?

From scheme-reports discussion list, by John Cowan:

Jeronimo Pellegrini scripsit:

According to Section 6.7.1, "Conversely, not all character ports are binary ports -- for example, the /string ports/ discussed below". It is not really clear to wether the document *requires* string ports not to be binary or if it was just an example of a port that *could* be character but not binary.

I haven't thought about it, but I guess it *could* be the latter, if the environment provides a default encoding for string ports.

  • Options: character-only, unspecified, undecided
  • Default: unspecified
  • Preferences: unspecified

Rationale: I see no reason to block an extension in this area.

#177 Distinguish file and string ports?

Should there exist predicates that identify string and file ports?

  • Options: string-port?, file-port?, both, neither, undecided
  • Default: neither
  • Preferences: neither

Rationale: I don't see the value of this.

#131 Output procedures return value

Output procedures (display, write, newline) currently return unspecified value, do we wish to make them return something (like in case of an error) or not?

Need proposals.

  • Options: r5rs, undecided
  • Default:
  • Preferences: r5rs

Rationale: There's not much you can do if these fail.

#134 Provide flush-output-port

This is an R6RS procedure that was part of PortsCowan, but never explicitly voted up or down by WG1. It flushes implementation output buffers on the specified port, the current output port by default.

  • Options: yes, no, undecided
  • Default: no
  • Preferences: yes

Rationale: Very helpful in TTY-style interactions such as student programs often use.

WG1 - Numerics

#117 Real numbers have imaginary part #e0

In R6RS, a complex number with imaginary part 0 is only real if the imaginary part is an exact 0. In R5RS, this was not true, and the requirement was simply that (zero? (imag-part Z)) be true.

  • Options: exact-only, any-zero, unspecified, undecided
  • Default: any-zero
  • Preferences: exact-only

Rationale: An inexact zero may or may not be zero: it can be any small real number whose absolute value lies between zero and the smallest positive real number representable as an inexact number. Numbers with inexact zeros are still complex, not real. This mildly breaks backward compatibility, but it does so in the direction of correctness.

#120 Define the semantics of the transcendental functions more fully

R6RS has an extended description of the transcendental functions. Do we want to include this?

TODO: explain the exact diff, why it is desirable, and whether any reasonable alternatives are possible.

References:

  • Options: r6rs, r5rs, undecided
  • Default: r5rs
  • Preferences: r6rs

Rationale: What it amounts to is that IEEE 754 semantics should be followed. We can just say that. In particular, R5RS has nothing to say about infinities and NaNs?, but we should for the benefit of implementations that support them (most).

#121 The semantics of expt for zero bases has been refined

This is a change also made by R6RS.

TODO: explain the exact diff, why it is desirable, and whether any reasonable alternatives are possible.

  • Options: r6rs, r5rs, undecided
  • Default: r5rs
  • Preferences: r6rs

#122 Make infinity, NaN, and -0.0 semantics (when supported) consistent with IEEE 754

R5RS does not explicitly describe these values. We have to decide whether to require that, if an implementation provides any of these values, they must be consistent with IEEE 754.

R6RS both requires these values and requires they be consistent with IEEE 754.

  • Options: ieee-754, unspecified, undecided
  • Default: unspecified
  • Preferences: ieee-754

Note: But only if supported by the implementation.

#175 Control of significant digits or decimal places in NUMBER->STRING

Vincent Manis pleads for a way to write numbers with a specified precision:

http://lists.scheme-reports.org/pipermail/scheme-reports/2011-May/000709.html

I (Alaric Snell-Pym) wondered if this should be done via NUMBER->STRING or via an optional extra argument to ROUND etc specifying a precision, as a number like 0.01 to get two decimal places. How to provide significant figures rather than DP without introducing a base-10 dependency is left as an exercise to the reader (as is the task of deciding if I'm mad for not wanting a base-10 dependency)

  • Options: manis, none, undecided
  • Default: none
  • Preferences: none

Rationale: Allowing rounded results in number->string mixes concerns.

#138 DivisionRiastradh domain

Zero as a divisor aside, what should the domain of the proposed procedures be?

  1. Any real numbers?
  2. Integers only?
  3. Exact integers only?
  • Options:
  • Default:
  • Preferences: integers

Rationale: I think division of non-integers is messy.

#217 DivisionRiastradh exactness preservation

What about exactness preservation?

  1. Not exactness preserving
  2. Exactness preserving unless the implementation can prove that an inexact argument can't affect the result (as in the case of an exact zero dividend and an inexact divisor)
  3. Exactness preserving in all cases
  • Options:
  • Default:
  • Preferences: exactness-preserving-unless

Rationale: This is consistent with everything else.

#140 Removing quotient, remainder, modulo

Are we removing the IEEE Scheme functions quotient, remainder, and modulo from WG1 Scheme? If so, we need a special justification, due to the charter text:

Existing features of IEEE Scheme may be removed only if a strong case can be made that they are fundamentally flawed. Insofar as practical, the language should be backwards compatible with the IEEE standard, the R5RS standard, and an appropriate subset of the R6RS standard.

Here's what DivisionRiastradh says:

Unfortunately, most programming languages give nondescript names such as DIV(IDE), QUOT(IENT), MOD(ULO), and REM(AINDER) to these operations. The language should make clear to programmers what division operations their programs are performing, especially when negative dividends and divisors can arise, but perhaps may not often arise during testing.

[...]

The R5RS gives the names quotient and remainder to the truncating division operator pair, and the name modulo to the remainder half of the flooring division operator pair. For all these three procedures in the R5RS, the dividend may be any integer, and the divisor may be any nonzero integer.

On the other hand, we may prefer relegating them to a backward-compatibility module.

Vote "yes" to keep, "no" to remove, and "module" to relegate to a module.

  • Options: yes, no, module, undecided
  • Default: yes
  • Preferences: yes

On reflection, every implementation will go on supporting them anyway, if just to keep old code working. Consistent with my "modules = optional" views, I see no reason to have a module here. There should be discouraging language in the report, though.

#151 Extend finite? and nan? to non-real values

R6RS specifies the domain of finite? and nan? as the real numbers only. I propose that finite? return #t on a non-real value iff both the real part and the imaginary part are finite and not +nan.0, and that nan? return #t on a non-real value iff either the real or the imaginary part is +nan.0.

  • Options:
  • Default:
  • Preferences: cowan

That is, what's proposed there. This is what people who do complex flonum programming need.

#152 exact-integer-sqrt inconsistent with multiple values module

R5RS does not actually specify any procedures which return multiple values, and so the decision to separate multiple values to a module was reasonable. However, we also voted to make exact-integer-sqrt, which is in the base module, return multiple values, namely the root and the remainder. That would make the procedure useless unless multiple values are provided.

We can either make multiple values not a module, make exact-integer-sqrt return a list (or single integer) rather than multiple values, relegate exact-integer-sqrt to a new module, remove it altogether, or do nothing and leave the inconsistency.

  • Options: values-in-core, return-list, return-pair, return-root-only, new-module, remove, nothing, undecided
  • Default: nothing
  • Preferences: values-in-core

Rationale: Multiple values are cheap to provide poorly for implementations that don't care, and can be provided well with some effort by implementations that do. No need to make them optional, therefore. Doing so also removes an extra dependency from WG2 packages that want to return multiple values.

#180 Make case and cond clauses into bodies

Andy Wingo suggests: make the clauses in case and cond forms (without =>, naturally) be BODY instances, to allow them to have definitions. It is well defined AFAIK, and costs nothing.

The counter-argument is that it doesn't "look" like the sort of place definitions are allowed.

  • Options: yes, no, undecided
  • Default: no
  • Preferences: no

Rationale: There is nothing in Scheme like ((= x y) (define z ...) (+ z x y)) except at top level. Let's not go there.

#181 Add WHEN and UNLESS to the base module

  • Options: yes, no, undecided
  • Default: no
  • Preferences: yes

Rationale: These are good things, and there is no reason to keep them separated out as if they were untouchables.

#182 Add WHILE and UNTIL

These trivial syntaxes add familiarity for new Scheme programmers coming from other languages, as will almost always be the case. LOOP is too big and named-LET too alien.

  • Options: yes, no, undecided
  • Default:
  • Preferences: yes

Rationale: Yes, I know we are not supposed to encourage mutability. But we have a lot of it already, particularly when dealing with ports.

#183 Escaped newline removes following whitespace?

Andy Wingo suggests the R6RS handling of escaped embedded newlines:

    "asdadf \
    asdfadf"

in R6RS has the same meaning as "asdf asdfadf". It allows you to nicely indent strings that you need to line-break for width. I suggest that the production

   \ NEWLINE WHITESPACE*

within string literals be elided.

Note an alternate method for handling embedded strings with nice indentation is scribble syntax.

We voted on various string syntaxes previously but did not specifically propose this R6RS extension. We should have a rationale if we don't follow it.

  • Options: yes, no, undecided
  • Default: no
  • Preferences: yes

Rationale: Cheap and useful.

#184 Require CHAR=?, STRING=? etc. to accept arbitrary numbers of arguments?

R5RS makes a point of specifying that supporting more than two arguments is optional. (Everything not explicitly mentioned is optional, so this may have significance.) R6RS requires accepting 2 or more arguments. Currently Racket, Gambit, Guile, Chez, Ikarus, Larceny, Ypsilon, Mosh, and Scheme 9 support the feature, whereas Gauche, MIT, Chicken, Bigloo, Scheme48/scsh, Kawa, SISC, Chibi, STklos, and SSCM don't.

  • Options: yes, no, undecided
  • Default:
  • Preferences: yes

Rationale: I think the R5RS implementations should be bootstrapped into the future here. It also meets the consistency guideline in the charter.

#185 Add sixth "centered" division operator

From the Guile manual:

— Scheme Procedure: centered/ x y — Scheme Procedure: centered-quotient x y — Scheme Procedure: centered-remainder x y

These procedures accept two real numbers x and y, where the divisor y must be non-zero. centered-quotient returns the integer q and centered-remainder returns the real number r such that x = q*y + r and -|y/2| <= r < |y/2|. centered/ returns both q and r, and is more efficient than computing each separately.

Note that centered-quotient returns x/y rounded to the nearest integer. When x/y lies exactly half-way between two integers, the tie is broken according to the sign of y. If y > 0, ties are rounded toward positive infinity, otherwise they are rounded toward negative infinity. This is a consequence of the requirement that -|y/2| <= r < |y/2|.

Note that these operators are equivalent to the R6RS operators div0, mod0, and div0-and-mod0.

--Andy Wingo

Taylor Campbell thinks these are useless. We should probably have use cases for _any_ division operator we include.

  • Options: yes, no, undecided
  • Default: no
  • Preferences: no

Rationale: I'm not convinced they are useful except maybe to do bignums with, in which case it's easy to define them.

Note: There are rationales for most of the division operators at Riastradh's original proposal.

#195 Editorial: proposed rewording for begin

The documentation for `begin' specifies that it is a sequential construct; but really it splices as well, and also of course it's a keyword for the module system currently. This is inaccurate of the spec to say that "begin is for sequencing".

Suggestion: adopt the language of R6RS section 11.4.7.

--Andy Wingo

We should explain somewhere the four kinds of begins: (begin expr ...), (begin decl ...), top-level begin, and module-top-level begin. Note that R7RS like R5RS does not have (begin decl ... expr ...).

Vote yes to adopt the R6RS description, modified for differences in the language.

  • Options: yes, no, undecided
  • Default: no
  • Preferences: yes

Rationale: I think this is a documentation issue only: the documentation Andy complained abuot is just for expression-begin.

#198 Make it an error for a procedure mapped by MAP and friends to mutate the result list/string/vector

This is possibly difficult to enforce, and can break existing R5RS programs written in very bad style.

  • Options: yes, no, undecided
  • Default: no
  • Preferences: yes

Rationale: I don't see how it can break anything, because no R5RS program can count on map using a mutable-result implementation anyway. Vector-map and string-map are more likely to.

#199 Make it an error for a procedure mapped by MAP and friends to return more than once

This is possibly difficult to enforce, and can break existing R5RS programs.

  • Options: yes, no, undecided
  • Default: no
  • Preferences: yes

Rationale: "Difficult to enforce" is irrelevant; "is an error" means no Scheme programmer should rely on the results of it.

#200 Completing the blob procedures

Add blob, blob-map, blob-for-each, and blob conversion functions to and from lists/vectors/strings.

  • Options: yes, no, undecided
  • Default: no
  • Preferences: no

Rationale: Too much for WG1.

#205 Roll partial-blob-copy(!) into blob-copy(!)

... with extra arguments.

  • Options: yes, no, undecided
  • Default: no
  • Preferences: no

Rationale: We did not roll substring into string-copy with additional arguments; the only reason not to use subblob as a name is its risibility.

#206 Provide read-syntax for blobs

R6RS provides a #vu8(...) read-syntax for bytevectors. SRFI-4 uses #u8(...).

  • Options: r6rs, srfi-4, none, undecided
  • Default: none
  • Preferences: none

Rationale: I can't see anyone writing these by hand. They might add a little efficiency compared to a blob procedure (analogous to string, vector, and list).

#207 Editorial: Polar complex numbers are inexact

Add a note saying that 1@2 and (make-polar 1 2) MAY evaluate to an inexact complex number.

  • Options:
  • Default:
  • Preferences: yes

Rationale: This is obviously an oversight: exact values passed to make-polar aren't going to come out exact from make-rectangular, and essentially all Schemes store complex numbers as rectangulars internally. (The Pure language supports both representations internally, but I can't see going there for standard -Scheme.)

#208 Is || a valid identifier?

The grammar in 7.1.1 allows || as an <identifier>. However, page 5 suggests the |...| form is only for convenience (e.g. |foo bar| is equivalent to foo\x20;bar). There's no way to normalise || to anything without the vertical bars that's a valid identifier. Was that intentional, or should the rule be

<vertical bar> <symbol element>+ <vertical bar>

Vote remove to remove the |...| syntax altogether.

  • Options: remove, empty-valid, empty-invalid, undecided
  • Default: empty-valid
  • Preferences: empty-valid
Rationale: Not sure why this is supposed to be the default.
is convenient on occasion, and allows string->symbol to accept any string.

#191 Include CLOSE-PORT ?

Should we include close-port, as a generic version of close-input-port and close-output-port?

  • Options: yes, no, undecided
  • Default: no
  • Preferences: no

Rationale: I could see adding this to a Scheme that has input/output ports to close both sides of the port, but the WG1 standard doesn't have them. Just as a generic procedure, no.

#188 Clarify wording of and and or definitions

The definitions of and and or may be slightly confusing. Reword them to be more clear. One possible hiccup is that the current language permits the return of different false values, while a clearer wording may preclude this.

R6RS provides a clearer definition that does not provide wiggle room for multiple false values. Should we use that?

  • Options: yes, no, undecided
  • Default: no
  • Preferences: yes

Rationale: I know why Guile needs multiple false values (to handle Emacs Lisp as well as Scheme), but I'd rather see such a system saying it overrides the Standard in this respect.

#187 Clarify duplicate bindings in let*

The language of the standard could clarify that duplicate bindings are permitted in the clauses of a let*.

  • Options: yes, no, undecided
  • Default: no
  • Preferences: yes

Rationale: Why not?

#215 initial value argument to make-blob

make-blob should either have an initial value argument, or rationale why it is inconsistent with make-vector and make-string.

Vote yes for an initial value argument.

  • Options: yes, no, undecided
  • Default: no
  • Preferences: yes

Rationale: As noted, I don't see blobs as specialized vectors. Still, I suppose initialization to all zeros couldn't hurt. Okay.

#216 Controlling use of reader labels on output

There are cases when one does not want to output reader labels for shared structure, such as when you don't care (and want the output to be more legible), or when you know that the time or space requirements to construct the table will be too large.

We could offer a parameter to control this, or have a separate procedure (e.g. write/simple) which doesn't use the reader labels.

Finer grained control may also let use specify a predicate for which values are interesting (e.g. never use labels for strings), or only use labels for cycles, etc.

  • Options: parameter, write/simple, none, undecided
  • Default: none
  • Preferences: write/simple

Rationale: I think a separate procedure is marginally better than a parameter.