Issue LOOP-MISCELLANEOUS-REPAIRS Writeup

Issue:        LOOP-MISCELLANEOUS-REPAIRS

Document:     (no document number yet)

Forum:        Cleanup

References:   refer to the referenced Public Review comments

Category:     CLARIFICATION

Edit history: 14-Feb-93, Version 1 by Moon

              18-Feb-93, Version 2 by Moon (fix spelling error)

Status:       Proposal FIX passed 7-1-0, March 1993


Problem Description:


  There were public review comments relating to chapter 6 of draft 12.24.

  This cleanup does not address the complete set, just the easier ones

  since it makes sense to lump them all into one omnibus issue.  I have

  not included editorial comments in this cleanup.


  A brief summary of the comments:


  1. Flanagan #4, Loosemore #3, Moon #47 -- extension mechanism is

  promised but not included


  2. Moon #28 -- text says "Function Designator", examples work differently


  3. Moon #30 -- LOOP "IT" Description doesn't make sense


  4. Moon #32 -- LOOP DOING doesn't allow implicit progn, but DO does


  5. Moon #35 -- LOOP-FINISH Inconsistency


  6. Moon #37 -- Does LOOP RETURN clause return multiple values?


  7. Moon #41 -- LOOP Conditional Execution Nonsense


  8. Moon #43 -- LOOP Bad Type-Spec Example


  9. Moon #21 -- LOOP Block Name Scope is not specified


  10. Moon #34 -- LOOP Thereis Default is not specified


  11. Moon #24 -- LOOP-AND-DISCREPANCY:NO-REITERATION not incorporated


  12. Moon #25 -- LOOP by/to phrase Order should be unrestricted


  13. Moon #36 -- LOOP Return/Named Interaction


  14. Moon #26 -- Eliminate the ability to put a LOOP keyword after FINALLY


Proposal (LOOP-MISCELLANEOUS-REPAIRS:FIX)


  Make all of the following numbered changes:

    

  1. Remove reference to extension mechanism on page 6-32.


  2. Change definition of step-fun on page 6-5 to:


    \param{step-fun}---a \term{form} that evaluates to a \term{function}

    of one \term{argument}.


  3. Replace the entire paragraph beginning "clause, clause1" on page 6-5

     with:


      \param{Clause}, \param{clause1}, \param{clause2} -- 

      \i{unconditional} | \i{accumulation} | \i{conditional}, as defined

      in the section "Syntax of an Extended LOOP Form".


     Replace the last paragraph on page 6-18 with: 


      The \term{loop keyword} \loopref{it} can be used to refer to the result

      of the test expression in a clause.

      Use the \term{loop keyword} \loopref{it} in place of the form in a

      {\tt return} clause or an \i{accumulation} clause that is

      inside a conditional execution clause.

      If multiple clauses are connected with \loopref{and}, the \loopref{it}

      construct must be in the first clause in the block.


    In a Notes section, note that:


      Use caution when using a variable named {\tt it} (in any package) in

      connection with LOOP, since \loopref{it} is a \term{loop keyword} that

      can be used in place of a form in certain contexts.


  4. Add DOING to the next-to-last paragraph on page 6-5.


  5. Remove the following paragraph from p.6-7:


      Note also that the \macref{loop-finish} macro terminates iteration and

      returns any accumulated result. Any \loopref{finally} clauses that are

      supplied are evaluated.


  6. Change the next-to-last paragraph on p.6-7, the third paragraph on

     p.6-19, and the last paragraph on p.6-19 all to state clearly that all

     values returned by the form in a RETURN clause are returned by the LOOP.


  7. Replace the first three paragraphs under the heading Conditional

     Execution on p.6-18 with:


       The \loopref{if}, \loopref{when}, and \loopref{unless} constructs

       establish conditional control in a \macref{loop}. If the test

       passes, the succeeding loop clause is executed. If the test does

       not pass, the succeeding clause is skipped, and program control

       moves to the clause that follows the \term{loop keyword}

       \loopref{else}. If the test does not pass and no \loopref{else}

       clause is supplied, control is transferred to the clause or

       construct following the entire conditional clause.


       If conditional clauses are nested, each \loopref{else} is paired

       with the closest preceding conditional clause that has no

       associated \loopref{else} or \loopref{end}. 


       In the \loopref{if} and \loopref{when} clauses, which are

       synonymous, the test passes if the value of \param{form} is

       \term{true}. 


       The \loopref{unless} \param{form} construct is equivalent to

       \f{when (not \param{form})}; the test passes if the value of

       \param{form} is \term{false}.


  8. Add OF-TYPE to the example on page 6-24.


  9. State explicitly in the next to last paragraph on page 6-19 that the

     scope of the block name includes the entire LOOP. Move the "BLOCK NIL"

     in the example on page 6-17 so it is outside the LET*.


  10. Add the following sentence to the description of "thereis" in the

      section "Termination Condition Clauses" on p.6-7:


       Otherwise, it provides a default return value of \nil.


  11. Make the syntax and the English agree with the X3J13 vote on

      LOOP-AND-DISCREPANCY:NO-REITERATION (Version 1) by correcting

      the syntax for each for-as subclause on page 6-3 by removing

      the "\curly{\loopref{for} | \loopref{as}}"; by adding the

      same text to the front of the definition of for-as-clause on

      page 6-3; and by adding a sentence to the second paragraph on

      page 6-9 clarifying that the word FOR or AS is not repeated

      after the word AND.


  12. The detailed description of for-as-arithmetic (p6-9) should say

      in English that the "by" subclause can appear either before or after

      the "to" subclause, and similarly for the other two pairs of

      subclauses. The syntax of for-as-arithmetic on p.6-3 should be

      changed as follows, to allow the three subclauses to appear in any

      order, to require at least one subclause to be used, and to indicate

      the restrictions on combinations of prepositions described in

      English on pp.6-9..6-10:


      for-as-arithmetic ::= { FOR | AS } var [ type-spec ]

       { { FROM | UPFROM } form1 [ { TO | UPTO | BELOW } form2 ] [ BY form3 ] 

|

         { FROM | UPFROM } form1 BY form3 [ { TO | UPTO | BELOW } form2 ] |

         { TO | UPTO | BELOW } form2 [ { FROM | UPFROM } form1 ] [ BY form3 ] 

|

         { TO | UPTO | BELOW } form2 BY form3 [ { FROM | UPFROM } form1 ] |

         BY form3 [ { FROM | UPFROM } form1 ] [ { TO | UPTO | BELOW } form2 ] 

|

         BY form3 { TO | UPTO | BELOW } form2 [ { FROM | UPFROM } form1 ] |

         FROM form1 { DOWNTO | ABOVE } form2 [ BY form3 ] |

         FROM form1 BY form3 { DOWNTO | ABOVE } form2 |

         { DOWNTO | ABOVE } form2 FROM form1 [ BY form3 ] |

         { DOWNTO | ABOVE } form2 BY form3 FROM form1 |

         BY form3 { DOWNTO | ABOVE } form2 FROM form1 |

         BY form3 FROM form1 { DOWNTO | ABOVE } form2 |

         DOWNFROM form1 [ { TO | DOWNTO | ABOVE } form2 ] [ BY form3 ] |

         DOWNFROM form1 BY form3 [ { TO | DOWNTO | ABOVE } form2 ] |

         { TO | DOWNTO | ABOVE } form2 DOWNFROM form1 [ BY form3 ] |

         { TO | DOWNTO | ABOVE } form2 BY form3 DOWNFROM form1 |

         BY form3 [ { TO | DOWNTO | ABOVE } form2 ] DOWNFROM form1 |

         BY form3 DOWNFROM form1 [ { TO | DOWNTO | ABOVE } form2 ]

       }


      Yes, it looks complicated, that's the limitation of BNF to express

      things where not all permutations are allowed.


  13. Replace the second sentence of the next-to-last paragraph on page 6-7

      with:


        It is equivalent to the clause

        \f{do (return-from \i{block-name} \i{value})},

        where \i{block-name} is the name specified in a \loopref{named}

        clause, or \nil if there is no \loopref{named} clause.


      Replace the second sentence of the third paragraph on page 6-19

      with the text quoted just above.


  14. Remove the syntax "finally unconditional-clause" on page 6-4.  Remove

      the sentence "An unconditional clause can also follow 

      \theloopkeyword{finally}." from the sixth paragraph on page 6-8.


Test Cases:


  Refer to the referenced Public Review comments


Rationale:


  These are non-controversial corrections to mistakes or ambiguities in

  the draft specification.


Current Practice:


  All of these changes bring the specification into conformity with current

  practice as observed in Apple MCL 2.0p1, with the exception of the ability

  to put the BY subclause before both of the other subclauses in #12.  I

  believe it would be a one-line change to the implementation to allow that.


Cost to Implementors:


  Little or none for most implementations that already conform.


Cost to Users:


  Probably none.


Cost of Non-Adoption:


  Ambiguous, inconsistent, or incomprehensible language specification.


Benefits:


  Better language specification.


Aesthetics:


  Better language specification.


Performance Impact:


  None.


Discussion:


  None.