Inform 7 v10's Compile to C question

can these instructions be followed in an IDE, or must it be through the command line?

It’s not laymans terms, but, I’m gonna do my best :stuck_out_tongue:

They’re for the command line; I don’t know what the IDEs support, sorry.

thank you anyway, I appreciate it.

I don’t believe the IDE apps are planning to support the C path. That will always be command-line only.

1 Like

From your inform/inform directory, create a dir for your games, eg games.
Then from inform/inform/games do:

../inform7/Tangled/inform7 -format=C -release mygame.i7
gcc -I../inform7/Tangled mygame.c -o mygame
1 Like

Is there an option to create C code to compile a standalone executable, or does additional interpreter code need to be added?

I think additional code would need to be added, @Wysardry.

Mr. @zarf has mentioned in this post: Inform 7 v10.1.0 is now open-source - #248 by zarf that this c-code is for putting Inform code into other frameworks like Unreal, not for creating executables.

That’s my ultimate goal, to code a game in Unreal or another engine using primarily Inform 7.

Yesterday I did manage to install Inform 7 to my mac’s terminal, and I got a game to ‘run’ and I think I’m very close to getting that c-output but I’m missing a step somewhere that turns it into an .i7 file first.

I followed Zed’s twitter instructions and he says that my c-code went to the build folder in the project, into the file auto.inf, and I opened that file up and it certainly looked like that had happened, so… I imagine just changing it from auto.inf to auto.c would work, but by then it was 4am and I needed to get some sleep.

My plan is then to stick the very simple, inform basic “hello world” into a game engine, and see if it runs.

Then… the real fun begins.

Inform’s C mode just produces a custom Glk app. There shouldn’t be any reason it couldn’t be compiled with GarGlk, Windows Glk, etc.

jkj yuio just posted how to do it.

You get a standalone executable – no additional code is necessary. You may link it with an external Glk library but by default it will use the built-in “Miniglk”.

I was thinking more along the lines of a DOS console program with no GUI, as I’d like to target retro computers.

I’ve never used Inform and the last time I used a C compiler directly was about 30 years ago, so a lot of info in the documentation (and in this thread) is lost on me.

1 Like

Glk is an I/O spec; it doesn’t imply a GUI. Miniglk only supports things that look like console programs.

But the C that’s produced requires a target that can handle 32-bit words. 16-bit or 8-bit targets could only be supported by compiling to Z-code and using a terp.

I would most likely be using SDCC which supposedly supports 8 to 64 bit and 4 byte data types according to their website.

Although i haven’t tried it, SDCC is likely not to work. I had huge problems compiling the C output. It definitely needs improving as it generates giant function that makes some compilers fall over. For example emscripten cannot compile the C as emitted. I had to break up functions into smaller pieces.


Thanks for letting me know. Can you keep us updated if the situation changes or you find a simpler solution?

I have other options I’m exploring, but compiling C code exported by Inform would likely be the simplest (if it worked).

A mid sized game generates about 16MB of source code into a single file. This is monstrous. The biggest culprit is

void i7_initialiser(i7process_t *proc);

This function alone can be 100,000 lines long. Compilers like to optimise on a function-by-function basis, and this one does have some things to improve it.

It’s clear Inform needs to do some work on the C code emitted. But i can understand the idea was to get it working first then optimise later. Which is the sensible way to go. Hopefully we’ll see an update that is much more optimised. For example, i think i7_initialiser could mostly be a large table.

Nevertheless, i think the C output is most exciting and the way to go for future Inform games.

Much as I dislike using C, I do agree that it was the best choice as the first programming language to export to.

Even if you ignore how many platforms you can get C compilers for, there are also plenty of C translator programs targetting Pascal, Nim, Go, WebAssembly, JavaScript etc.

So I have compiled my very simple “Hello.i7” Which I basically copied from the eg-1:

It has compiled to c with over 86,000 lines of code, turning my 32bit program into a 3mb program, mostly of definitions. I used the -basic command, so I thought it wouldn’t use those, but I guess it’s something that inform does need to translate the english into something the computer will understand.

I did get it to print “hello world” in VS code after a little finagling with paths to inform_clib.
I think I can get it to print to Unreal’s console…

After that I don’t know what to do next, since I am still learning how to make games outside of interactive fiction.

Even without the Parser/World Model, there’s a lot of stuff built into Basic Inform: you have all the infrastructure for relations, tables, rules, rulebooks, activities, and generating adaptive text. For instance, you could write this:

Datum is a kind of object.
The plural of datum is data.

Surveys is a plural-named datum.
Graph is a datum.

To demonstrate is a verb.

To begin:
  say "[The surveys] [demonstrate].";
  now the story tense is future tense;
  say "[The graph] [demonstrate]."

And there’s no attempt to strip unused functions. And the way that Inter is translated into C seems to work a lot like: imagine what glulx bytecode the I6 compiler would generate for this bit of Inter, and then generate C that imitates each of those bytecodes in turn.

With gcc, the program above compiled to a 985K binary, but letting gcc optimize for size and strip unused stuff, it got down to 384K.

gcc -lm -I"inform7/Tangled"  -Os -flto -fdata-sections -ffunction-sections basic.c -o basic  -Wl,--gc-sections

The C output is a great start, but it does need some work. I do think, it could be a lot better. But i understand the first priority is always to get it working.