annotate admin/notes/iftc @ 83509:24cf4bf418dc

Merged from [e-mail] Patches applied: * [e-mail]/emacs--devo--0--patch-207 Update from CVS * [e-mail]/emacs--devo--0--patch-208 Sync from erc--emacs--0 * [e-mail]/emacs--devo--0--patch-209 Update from CVS * [e-mail]/emacs--devo--0--patch-210 Merge from gnus--rel--5.10 * [e-mail]/emacs--devo--0--patch-211 Update from CVS * [e-mail]/emacs--devo--0--patch-212 Update from CVS * [e-mail]/emacs--devo--0--patch-213 Update from CVS * [e-mail]/emacs--devo--0--patch-214 Merge from gnus--rel--5.10 * [e-mail]/emacs--devo--0--patch-215 Update from CVS * [e-mail]/gnus--rel--5.10--patch-81 Update from CVS * [e-mail]/gnus--rel--5.10--patch-82 Update from CVS * [e-mail]/gnus--rel--5.10--patch-83 Update from CVS git-archimport-id: [e-mail]--2004/emacs--multi-tty--0--patch-549

Iso-Functional Type Contour


This is a term coined to describe "column int->float" change approach, and can
be used whenever low-level types need to change (hopefully not often!) but the
meanings of the values (whose type has changed) do not.

The premise is that changing a low-level type potentially means lots of code
needs to be changed as well, and the question is how to do this incrementally,
which is the preferred way to change things.

Say LOW and HIGH are C functions:

  int LOW (void) { return 1; }
  void HIGH (void) { int value = LOW (); }

We want to convert LOW to return float, so we cast HIGH usage:

  float LOW (void) { return 1.0; }
  void HIGH (void) { int value = (int) LOW (); }  /* iftc */

The comment /* iftc */ is used to mark this type of casting to differentiate
it from other casting.  We commit the changes and can now go about modifying
LOW and HIGH separately.  When HIGH is ready to handle the type change, the
cast can be removed.

;;; arch-tag: 3309cc41-5d59-421b-b7be-c94b04083bb5