To: J3                                                     J3/18-156
From: Steve Lionel & Bill Long & Dan Nagle & Lorri Menard
Subject: J3 initial recommendations for F202X features
Date: 2018-February-26

Reference: N2147, 18-122r1, minutes215.txt

Introduction
------------

At its February 2018 meeting, J3 made a first pass over a long list of
suggestions for new features in the next Fortran standard revision
(working title Fortran 202X). The user survey run since last summer,
and summarized in WG5 document N2147, provided the bulk of suggestions,
along with relative ranking of selected features. Some J3 members
offered their own suggestions as well.

From the user survey, as well as general sentiment on the committee,
it was clear that two major features were considered to be of very
high value:

- Generic programming or templates
- Exceptions

Multiple papers on these ideas have been submitted over the years -
recent instances are 18-108, 18-109, 18-112, 18-115, and 18-116.

J3 members in attendance were asked to submit a list of up to five
features, other than generics and exceptions, with use cases where
applicable. These were compiled into paper 18-122r1. The items were
divided among the Data, HPC and JoR subgroups; the subgroups
considered each suggestion and recommended priorities for each. J3
as a whole then met and took straw votes on each item as to whether
to recommend further development of the feature. This paper details
the suggestions recommended for consideration by WG5. Note that
feature specifics (syntax, etc.) are generally left unspecified; if
a feature is accepted by WG5 details will be developed by J3 from
use-cases and the resulting feature may vary from any specific
proposal here.

A brief note regarding the popular suggestion of "containers" (see
N2147, 18-110r1, 18-123) - there was general agreement the position
expressed in 18-142, "Data subgroup believes that it would be better
to have standard features in Fortran that will allow developers to
create general or specific container objects.", was the preferred
direction.

Recommended Feature List
------------------------

J3 recommends to WG5 that the following features be included in
Fortran 202X. References to J3 papers are given where available.
Numbers in parentheses, such as (3), indicate support from multiple
members. Straw votes, where taken, are shown.

Major features

Generic programming or templates
   No straw vote: Forward 18-142 as J3 recommendation on Containers
   Note that Generics from Van Snyder (Ada feature) and Templates
   (modeled on C++) are quite different.
   18-108 Generic (Clune)
   18-112 Generic (Snyder)
   18-116 Templates (Long)

Exceptions
   No straw vote.
   18-109 Error Handling (Clune)
   18-115 Exceptions (Van)
   94-254r4 Exceptions (Reid)
   BCS17b.pdf Exceptions (Reid)

Approved for more work:
-----------------------

Medium/Large - Bit strings
  (Y:8, N:0, U:4)
  In general, the BITS feature that was stripped from F2008 toward
  the end of development. Needs a paper with old design citations and
  use cases. Could replace unsigned integers, and also solve the
  real/complex problem (see 18-128).

Medium/Large - New types, not type aliases, from existing types
  (Y:8, N:1, U:4)
  18-113r1

Medium/Large - Proper enumerated types; Type system affected. Possibly a
   bind(c) feature? (Y:6, N:0, U:4)
   18-114r1

Medium - Asynchronous subroutine execution with event post on
   completion
   (Y:8, N:0, U:5)
   18-125
   Work to do:
   Need alternative to EVENT WAIT.
   Race conditions - need edits like ASYNCHRONOUS for I/O
   Van wants other models for this more like Ada.
   How related to OMP tasks?

Medium - (2)An ability to declare a reduction variable on DO
   CONCURRENT.
   (Y:7, N:0, U:3)
   18-143

Medium - Put with Notify
   (Y:6, N:0, U:6)
   18-117r1

Medium/Small - (3)Specifying default values for optional dummy
  arguments that are omitted
  (Y:6, N:4, U:2)
  18-136r1
  Extensive discussion about unwanted effects on scoping rules and
  possible conflicts with SAVE. Most agreed that if the caller provided
  an anonymous writable temporary, in the same manner as VALUE for
  non-BIND(C) procedures, that most if not all problems would be avoided.
  For syntax consider =value in the dummy argument list, rather than
  overloading initializers, though there is no semantic ambiguity.

Medium/Small - Protected structure components (Y:7, N:0, U:3)
   18-146

Small - User control of the optional leading zero in F/E format
   (18-120)
   (Y:9, N:0, U:4)
   18-120r3

Small - Degree trigonometric functions
   (Y:9, N:0, U:1)
   18-139r1
   Already implemented by several vendors.

Small - Significantly increase free-form line length
   (Y:10, N:0, U:3)
   18-124
   Work to do:
   Refocus on allowing longer total statement length with more
   continuation lines an option.

Small - A0 format that trims whitespace from right of strings
   (Y:7, N:0, U:6)
   18-126r1

Small - Short-circuit logical operators
   (Y:8, N:0, U:2)
   18-152

Small - More CHARACTER intrinsics (delimiter processing, perhaps?)
   (Y:7, N:1, U:5)
   18-127
   Work to do:
   Concern about name conflicts with existing user procedures
   Aimed at deletion of Part 2.
   List needs to be narrowed, removing more trivial functions.

Small - Allow deferred-length allocatable char vars for messages,
  intrinsics that return strings
  (Y:10, N:2, U:1) for errmsg and iomsg specifiers
  (Y:7: N:0, U:6) for intrinsics that return strings other than error
  messages
  (Y:7, N:0, U:6) for write to internal files
  18-130

Small- A VIRTUOUS (superpure) procedure prefix that declares the
  procedure does not access data outside of the procedure. (Y:5, N:0,
  U:5)
  18-151

Small - Syntax to specify that the target of a dummy argument
  pointer is not to be changed (INTENT applies to the pointer, not its
  target) (Y:7, N:0, U:3)
  18-144

Small - An INTENT syntax that means the same as not specifying
   INTENT
   (Y:7, N:2, U:3)
   18-150

Small -  add LOG8, LOG16, LOG32, LOG64, and REAL16 to
   iso_fortran_env
   (Y:10, N:0, U:2)
   18-148
   Originally: Remove requirement that LOGICAL requires one numeric
   storage unit

Small - CFI function to return text version of a processor's CFI
  error code
  (Y:10, N:0, U:2)
  18-145

Small - Binary compatibility of C descriptors so that a single C
  library could work with multiple Fortran implementations
  (Y:11, N:0, U:1)
  Needs more work and a solid design. Concept popular, though. Maybe look
  at additional CFI routines rather than breaking descriptor compatibility

Small - Providing methods for easily accepting NUL-terminated
  strings in calls to Fortran from C.
  (Y:7, N:0, U:5)
  18-147
  Current design involves conversion intrinsics.

Small - POINTERs initially null instead of undefined (Y:6, N:3, U:1)
   18-140

Undecided:
----------

Undecided, Small - += and *= syntax
   No paper or Vote. JoR rated this low priority.

Undecided - Constructors - procedure called automatically when a
   derived type comes into existence. No vote or paper.

Undecided - Allowing a specification section in (some) constructs
  without forcing  use of BLOCK/END BLOCK seems useful.
  No paper. Defer discussion to next meeting.

Undecided - Allow arrays and allocatables with coarray
   components. Continue discussion at the next meeting. No vote.
   17-200r1


--END--