To: J3                                                     J3/21-197
From: Steve Lionel
Subject: Changeable modes and derived type I/O
Date: 2021-October-20

Reference: 21-007r2

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

Van Snyder's paper 21-177 raises the question of just what the standard
says should be the value of the scale factor for a child data transfer
statement in derived type formatted I/O. The issue is broader in that the
same question applies to all of the "changeable modes" (12.5.2). For
example, if the parent transfer has a 3P edit descriptor, is that in
effect for the child transfer? And, what happens if the child transfer
has its own control edit descriptor or keyword to change one or more of
the modes - does this change flow back to the parent upon completion?

A related question is whether the connection of the child transfer is the
same as that of the parent. This is plausibly the case for external I/O,
but for internal I/O, the child is supplied a negative unit number and
not the character variable used in the parent - is this really the same
connection?

12.5.2p2 says:
"Values for the modes of a connection are established when the connection
is initiated. If the connection is initiated by an OPEN statement, the
values are as specified, either explicitly or implicitly, by the OPEN
statement. If the connection is initiated other than by an OPEN statement
(that is, if the file is an internal file or preconnected file) the values
established are those that would be implied by an initial OPEN statement
without the corresponding keywords."

The concensus of JOR is that changeable modes are intended to "flow in"
to a child transfer, but not flow back out. There is note 12.1 that says,
in part:

"The effect of executing thedefined input/output procedure is similar to
that of substituting the list items from any child data transfer
statements into the parent data transfer statement's list items, along
with similar substitutions in the format specification."

If one takes this literally, then mode changes flow in and out of child
transfers, which seems undesirable. A test of four current compilers
shows that three do not inherit modes from the parent and one does. None
of the four have mode changes flow back out to the parent. However, the
note does say "is similar to", so perhaps it is not intended to be exact.

Proposal
--------

Specify that, upon initiation of a defined input/output procedure, the
unit specified by the UNIT dummy argument identifies a temporary copy
of the parent's file connection, including the current values of all
changeable modes. 12.5.2p5 already says "When a data transfer
statement terminates, the values for the modes are reset to the values
in effect immediately before the data transfer statement was executed."

Note that one is allowed to use INQUIRE to ask about the changeable modes
other than scale factor. (See note 12.6 on p253.)

Edits to 21-007r2
-----------------

230:23-26 12.5.2p2 "Connection Modes"

Replace the third sentence ("If the connection is initiated other than
by an OPEN statement") with:
"If a defined input/output procedure (12.6.4.8.2) is called, the values
for the connection identified by the *unit* dummy argument are those
presently in effect in the parent data transfer statement. Otherwise
(that is, if the file is an internal file or preconnected file) the
values established are those that would be implied by an initial OPEN
statement without the corresponding keywords."

--END--