inform7/resources/Changes/Change Logs/3V01.txt

3V01 (2 August 2006)

Rather than being a bug-fixing build, this build carries out a number of
	significant internal reforms, some outwardly visible and some not.
For the first time, compilation to the Glulx virtual machine rather than
	the Z-machine is possible. As the option on the Settings panel for a
	project suggests, we had always intended this to be a possibility,
	but too much was involved to get Glulx working in time for the public
	beta. In the future, we hope that Glulx will provide a solidly
	reliable platform for larger and for multimedia works: at present,
	though, our aim is simply to allow larger works to be compiled:
	and there are a few restrictions on screen effects, with some code
	(particularly to do with coloured text) not yet working.
	Glulx support in this build is experimental at best: we ask users
	not to send bug report forms yet, and would advise that people avoid
	publishing works compiled to Glulx rather than Z. With that said, we
	do believe that worthwhile progress is being made.
The most important change in adapting to Glulx was the switch to an
	Inform 6 library derived from library 6/11, the current biplatform
	I6 library: up until now, Inform 7 has used a heavily adapted I6
	library referred to as "6/10N", derived from both 6/9 and 6/10.
	The new library, "6/11N", is the work of Jesse McGrew: our intention,
	once Glulx support is stabilised a little, is that this will
	eventually become library 6/12 - that is, the standard I6 library
	for use by both I6 and I7 authors alike. This should finally bring
	together branches of work on the I6 library which have been divergent
	for three years now.
	For the most part, the change of library will be invisible to I7
	users, but the adoption of various I6 bug-fixes and improvements
	between 6/9 and 6/11 should subtly improve the behaviour of Inform
	story files at run-time.
Inform now reacts more helpfully to various overflows in Inform 6: thus, if
	Inform generates code which overflows the size of the Z-machine or
	breaks one of Inform 6's memory settings, rather more useful pages of
	advice are displayed.
As promised, the deprecated "means" and "means that" features have been
	withdrawn, and removed from Chapter 19 of the documentation.
The meanings of "in" and "on" have been made stricter: in past builds, the
	test "if A is in B" would pass if, e.g., A was being carried by a person
	B: this now fails. "A is in B" requires B to be a room or container;
	or else requires A to be a room or region, and B a region. Similarly,
	"A is on B" now requires B to be a supporter. The point of these reforms
	is that the built-in relations (well, all relations, really) ought to
	test as the same relation that they assert: asserting "The spanner is
	in the box." implies that "the box" is definitely not a person, so it
	is wrong for "if the spanner is in Fred" to succeed where Fred clearly
	is a person.
Because it is, nevertheless, useful to be able to test "by whatever means
	A is immediately carried by B", the meaning of "to hold" has been
	changed to this relation. This accords with the meaning of the existing
	construction "the holder of A": thus "B holds A" if and only if B is
	"the holder of" A, and one can talk about "things held by B", and so on.
	In previous builds, "to hold" was synonymous with "to carry", and
	implied that the holder was a person: clearly, this is no longer true.
	A bucket can "hold" sand, a table can "hold" place settings, and so on.
	Note that, as with "holder of", the parts of something are also deemed
	to be "held" by it.
Examples:
	Minor changes have been made to a small number of examples to reflect
		this stricter sense of "in";
	Two new examples, "Bees" and "Zqlran", demonstrate a random maze and
		and an exotic time notation respectively.
Bug fixed so that "now" is able to change variables and properties, e.g. by
	writing "now the score is 132" or "now the carrying capacity of the
	player is 10". (In general, the "now" mechanism for asserting things
	has been rewritten, and should be more capable all round.)
When Inform deals with descriptions such as "the people who are in lighted
	rooms", it compiles code using Inform 6's "objectloop" construct. The
	running speed of the result depends in part on the form of objectloop
	used, and in general it is fastest to try to minimise the size of the
	set of the objects looped through. Inform 7 now does a better job of
	minimising this set, and tries to avoid circumstances in which it is
	forced to loop through every object in searching for a match to a
	pattern. (It is possible that, in a limited number of cases, a
	description such as "if a woman (called the shopper) is in a shop, ..."
	will now result in a different woman being written into the "shopper"
	value than would have happened under previous builds - still a
	woman who is in a shop, but maybe not the same one. This might possibly
	affect Transcripts being used for testing.)
	Because this optimisation trades memory space for time, it is switched
	off if "use memory economy" has been requested.
Values of kind "rule" can now be said - the result being to print their
	names: e.g., "can't turn scenery rule". This will probably only be
	useful for debugging, or the writing of surrealist competition pieces.
The scene icons missing in build 3T38 for OS X have been restored.
The crash occurring in build 3T38 for Windows to do with "a, b, and c" has
	been fixed.
Bug in listing installed extension documentation fixed.
Problem messages in the EPS map-maker generally tidied up.