mirror/ioccc
1
0 Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Format and typo fixes in 2019/adamovsky/README.md

Browse Source
This commit is contained in:
Cody Boone Ferguson committed 2023-04-02 12:00:55 -07:00
1 parent 0c543beb0f
commit e0a9b41a32
1 file changed
+128 -121
Download Patch File Download Diff File Expand all files Collapse all files
2019/adamovsky/README.md
+128 -121
View File
@@ -1,6 +1,6 @@
# Most functional interpreter # Most functional interpreter
Ondřej Adamovský <oa@cmail.cz> Ondřej Adamovský <oa@cmail.cz>
## To build: ## To build:
@@ -29,9 +29,9 @@ make
Even after pre-processing and indenting, the C code of this entry Even after pre-processing and indenting, the C code of this entry
is about as understandable as the Unlambda code. is about as understandable as the Unlambda code.
Pretending that we don't know a bit of functional programming :) and lacking a better way Pretending that we don't know a bit of functional programming :) and lacking a
to understand the entry but to fuzz it, better way to understand the entry but to fuzz it, we stumbled on a string of
we stumbled on a string of bytes which crashed it: bytes which crashed it:
``` ```
````.0`.0`.0`c.0``sssss ````.0`.0`.0`c.0``sssss
@@ -43,38 +43,40 @@ Functional programming is not a panacea against core dumps, after all.
**&#187;A highly structured state state machine&#171;** **&#187;A highly structured state state machine&#171;**
### Purpose ### Purpose:
The purpose of this program is to allow you to play Colossal Cave Adventure as The purpose of this program is to allow you to play Colossal Cave Adventure as
implemented by [Kunihiko Sakamoto][1]. As that was written in [Unlambda programming implemented by [Kunihiko Sakamoto][1]. As that was written in [Unlambda programming
language][2] designed by David Alexander Madore it has to be an Unlambda interpreter. language][2] designed by David Alexander Madore it has to be an Unlambda interpreter.
### Aesthetics ### Aesthetics:
The program grew a bit too long so I had to use several macros to downsize it. The program grew a bit too long so I had to use several macros to downsize it.
That unfortunately reduced my options for program layout. I decided to separate That unfortunately reduced my options for program layout. I decided to separate
it into two portions. The initial macro section where I had limited resources and it into two portions. The initial macro section where I had limited resources
you should disregard its shape, unless you like it, and the rest is a separator in the and you should disregard its shape, unless you like it, and the rest is a
form of a single line followed by a section signifying the function of the program, separator in the form of a single line followed by a section signifying the
a block with removed shape of lambda character. Even the official IOCCC size tool function of the program, a block with removed shape of lambda character. Even
acknowledges this: When you call `./iocccsize -i <prog.c`, using the the official IOCCC size tool acknowledges this: When you call `./iocccsize -i
[2019 version of iocccsize](https://www.ioccc.org/2019/iocccsize.c), <prog.c`, using the [2019 version of
iocccsize returns **955**, which is the Unicode codepoint of the **&#955;** character. iocccsize](https://www.ioccc.org/2019/iocccsize.c), iocccsize returns **955**,
which is the Unicode codepoint of the **&#955;** character.
It might be a surprise for you that according to the IOCCC size tool the program is It might be a surprise for you that according to the IOCCC size tool the program is
actually a *oneliner*. It certainly was for me. I like to think the tool is overwhelmed actually a `one-liner`. It certainly was for me. I like to think the tool is overwhelmed
by the sheer length of the dividing line and totally overlooks the rest of the program. by the sheer length of the dividing line and totally overlooks the rest of the program.
### Usage ### Usage:
When built directly When built directly using
```sh ```sh
make make
``` ```
the program (*prog*) accepts a single parameter with the name of an Unlambda program. You can the program (`prog`) accepts a single parameter with the name of an Unlambda program. You can
download the [Colossal Cave Adventure][1] (the *advent.unl* file) and run it like this download the [Colossal Cave Adventure][1] (the [advent.unl](advent.unl) file)
and run it like this:
```sh ```sh
./prog advent.unl ./prog advent.unl
@@ -84,24 +86,24 @@ You can also try other programs from the Web. Many example programs are in the o
[Unlambda distribution][3] on the Unlambda homepage. The most complex programs there are [Unlambda distribution][3] on the Unlambda homepage. The most complex programs there are
entries from [quine contest][4]. entries from [quine contest][4].
There is an alternative build path that requires IOCCC size tool to complete. Put the tool There is an alternative build path that requires the 2018 IOCCC size tool to complete. Put the tool
source *iocccsize.c* in the project directory and call source [iocccsize_2018.c](iocccsize_2018.c) in the project directory and call
```sh ```sh
make identify make identify
``` ```
It will build the tool and use it to build an alternative program *prog2*, which in turn will It will build the tool and use it to build an alternative program `prog2`, which in turn will
produce an Unlambda program *identify.unl*. `make identify` runs this program to identify produce an Unlambda program `identify.unl`. `make identify` runs this program to identify
itself. itself.
The base program (*prog*) implements an interpreter of Unlambda 2.0. It recognises all its The base program (`prog`) implements an interpreter of Unlambda 2.0. It recognises all its
builtins (k, s, i, v, c, d, .*x*, e, @, ?*x* and |) including the syntactic sugar function builtins (`k`, `s`, `i`, `v`, `c`, `d`, `.x`, `e`, `@`, `?x` and `|`) including the syntactic sugar function
(&#8203;r&#8203;) and formating properties in the form of ignoring whitespace and comments `r` and formatting properties in the form of ignoring whitespace and comments
to the end of line introduced by the *#* character. to the end of line introduced by the `#` character.
In case of wrong usage (missing Unlambda program file or missing parameter) or use of an unseekable In case of wrong usage (missing Unlambda program file or missing parameter) or use of an unseekable
Unlambda program source if it uses the *c* function, the program returns message *fail*. It Unlambda program source if it uses the `c` function, the program returns message `fail`. It
also informs you about syntactic errors in the Unlambda program should there be any (but only also informs you about syntactic errors in the Unlambda program should there be any (but only
when the execution reaches them). when the execution reaches them).
@@ -111,37 +113,37 @@ them all in sequence.
### Unlambda ### Unlambda
If you want to familiarise yourself with the language, visit its [Wikipedia page][5] for If you want to familiarise yourself with the language, visit its [Wikipedia page][5] for
a brief introduction or its [homepage][2] for an indepth discussion. For further reading you a brief introduction or its [homepage][2] for an in depth discussion. For further reading you
can visit Wikipedia pages on [Lambda Calculus][6] and [Combinatory Logic][7]. can visit Wikipedia pages on [Lambda Calculus][6] and [Combinatory Logic][7].
You can also play with an [online interpreter][8] by Github user inazz. There you can You can also play with an [online interpreter][8] by GitHub user inazz. There you can
debug one of example programs provided or try your wits in writing your own. debug one of example programs provided or try your wits in writing your own.
### Hints ### Hints:
Here are some pointers where to start when trying to understand the interpreter: Here are some pointers where to start when trying to understand the interpreter:
- Where is syntax error detected? - Where is syntax error detected?
- How is the memory managed? - How is the memory managed?
- How does a program character turn into a function? - How does a program character turn into a function?
- How does exit function *e* work? - How does exit function *e* work?
- Can fD0 function be called? - Can fD0 function be called?
- Are the function names familiar? (warning: some are misleading) - Are the function names familiar? (warning: some are misleading)
### Obfuscation sources ### Obfuscation sources
- The elephant in the room is the Unlambda language. As we all know, functional - The elephant in the room is the Unlambda language. As we all know, functional
programming defies sanity. Furthermore, Unlambda was specifically designed programming defies sanity. Furthermore, Unlambda was specifically designed to be
to be unintelligible and hard to code in. Most of its functions are quite unintelligible and hard to code in. Most of its functions are quite simple and
simple and straightforward to implement, but two make it a mess: the promise straightforward to implement, but two make it a mess: the promise operator `d`
operator *d* and the call with current continuation *c*. and the call with current continuation `c`.
- Most of the additional obfuscation is derived from basic design choices. That - Most of the additional obfuscation is derived from basic design choices. That
is, how the state of the program is stored and how are the Unlambda functions is, how the state of the program is stored and how are the Unlambda functions
implemented. implemented.
More on this in the Spoilers section. More on this in the **Spoilers** section.
### Build notes ### Build notes:
The program does not require any special treatment by compilers. Some warnings The program does not require any special treatment by compilers. Some warnings
had to be suppressed when compiling with `-Wall -Wextra -pedantic`. Those were had to be suppressed when compiling with `-Wall -Wextra -pedantic`. Those were
@@ -150,105 +152,110 @@ technique. Compilation with `-Weverything` produced many more warnings, but thos
are only relevant for multi-source-file projects and some are really questionable are only relevant for multi-source-file projects and some are really questionable
(e.g. `-Wdisabled-macro-expansion`). (e.g. `-Wdisabled-macro-expansion`).
### Portability ### Portability:
Development was done on Debian 9 on x86-64 architecture using mainly Clang 3.8.1-24 Development was done on Debian 9 on x86-64 architecture using mainly Clang 3.8.1-24
and GCC 6.3.0. The source is C11 compliant and does not use any special properties and GCC 6.3.0. The source is C11 compliant and does not use any special properties
except the following: except the following:
- The interpreter sort of relies on tail call optimization. It works well even - The interpreter sort of relies on tail call optimization. It works well even
without it, but the stack runs out quickly with some Unlambda programs (e.g. *trivial.unl* without it, but the stack runs out quickly with some Unlambda programs (e.g.
example from the [reference Unlambda distribution][3] produces segfault). `trivial.unl` example from the [reference Unlambda distribution][3] produces
Clang with `-O3` option does this optimization well and said example does segfault). Clang with `-O3` option does this optimization well and said example
not segfault with it. I was not able to force GCC to optimize it out (probably does not segfault with it. I was not able to force GCC to optimize it out
an operator fault). (probably an operator fault).
- The only data structure of the program (the union structure) had much more - The only data structure of the program (the union structure) had much more
members originally. I joined all integers to a single long long member in the name members originally. I joined all integers to a single long long member in the
of obfuscation. It is used for reference counting (only 1 for every reference) among name of obfuscation. It is used for reference counting (only 1 for every
other uses. size_t should be used for that, but practically, I think the memory reference) among other uses. `size_t` should be used for that, but practically,
would run out sooner than the reference counter in any real world situation I think the memory would run out sooner than the reference counter in any real
even when size_t is bigger than long long. It is also used for storing chars world situation even when `size_t` is bigger than `long long`. It is also used
for later printing. This might produce some warnings, but since it was developed for storing chars for later printing. This might produce some warnings, but
using chars it should be always used safely. since it was developed using chars it should be always used safely.
### Spoilers ### Spoilers:
- You can examine the *progSpoiler.c* file. It is the original final version of - You can examine the [progSpoiler.c](progSpoiler.c) file. It is the original
the interpreter, before I started the downsizing. It accepts one more optional final version of the interpreter, before I started the downsizing. It accepts
parameter that specifies filename for the output Unlambda function (it's the one more optional parameter that specifies filename for the output Unlambda
Unlambda function, that remains after all function applications are done). function (it's the Unlambda function, that remains after all function
When no filename is provided, *stderr* is used. applications are done). When no filename is provided, `stderr` is used.
The function *output* is used for this additional functionality. You can use The function *output* is used for this additional functionality. You can use
it while debugging to visualise current variables. it while debugging to visualise current variables.
- The Unlambda functions are represented by a record consisting of an array of five - The Unlambda functions are represented by a record consisting of an array of
multipurpose arguments. The first one refers to the function implementing its five multipurpose arguments. The first one refers to the function implementing
application, the second and the third ones are for arguments of the function, its application, the second and the third ones are for arguments of the
the fourth one is for stack building, and the fifth one is the reference counter. function, the fourth one is for stack building, and the fifth one is the
reference counter.
The function argument is used not only for evaluation of the functions, but also The function argument is used not only for evaluation of the functions, but also
as a state variable to decide whether we want to call it, or do something else. as a state variable to decide whether we want to call it, or do something else.
- The interpreter reads the Unlambda source as it interprets it, translating - The interpreter reads the Unlambda source as it interprets it, translating the
the incoming character to a function using a constructor stored in translation array *T*. incoming character to a function using a constructor stored in translation array
`T`.
This means the interpreter must reposition in the source file whenever current This means the interpreter must reposition in the source file whenever current
continuation is called. continuation is called.
- The state of the interpreter is represented by trees representing evaluated - The state of the interpreter is represented by trees representing evaluated
Unlambda functions stored on the C stack as local variables, plus the source file Unlambda functions stored on the C stack as local variables, plus the source
position. When a current continuation is created, the stack is unwound using the file position. When a current continuation is created, the stack is unwound
fourth argument of the nodes, stored in the continuation, and rebuilt back resuming using the fourth argument of the nodes, stored in the continuation, and rebuilt
the evaluation. When its called, the stack is unwound and thrown away, and the back resuming the evaluation. When its called, the stack is unwound and thrown
stack stored in the continuation is restored. away, and the stack stored in the continuation is restored.
This makes implementation of a complete debugger a bit more challenging. This makes implementation of a complete debugger a bit more challenging.
For this to work, the functions that do the work (i.e. make series of applications For this to work, the functions that do the work (i.e. make series of applications
and/or evaluations each of which may be a call to current continuation), the *s* and/or evaluations each of which may be a call to current continuation), the `s`
combinator and the *`* application operation had to be broken down to pieces, one combinator and the `` ` `` application operation had to be broken down to pieces, one
for each application/evaluation and corresponding helper functions for stack for each application/evaluation and corresponding helper functions for stack
rebuilding. (the helper functions are named beginning with *r* instead of *f*) rebuilding. (the helper functions are named beginning with *r* instead of *f*)
- The *`* application operation does the reading of the Unlambda source when needed. It - The `` ` `` application operation does the reading of the Unlambda source when
also implements the promise function. Whenever it would apply it, it overwrites needed. It also implements the promise function. Whenever it would apply it, it
the Apply constructor in translation array *T*, changing its function to build overwrites the Apply constructor in translation array `T`, changing its function
unevaluated subtree that is later evaluated whenever the promise is applied. to build unevaluated subtree that is later evaluated whenever the promise is
applied.
- To further the obfuscation, I shortened the function names while downsizing, - To further the obfuscation, I shortened the function names while downsizing,
but preserved their meaning (with the exception of Apply functions, which I but preserved their meaning (with the exception of Apply functions, which I
renamed to Y to possibly confuse with Y combinator). Simple functions retained renamed to Y to possibly confuse with Y combinator). Simple functions retained
their names to misguide the reader to believe to know what is what. The names their names to misguide the reader to believe to know what is what. The names of
of continuation and exit functions were swapped. The names of S and Apply continuation and exit functions were swapped. The names of S and Apply functions
functions were shuffled both for S functions and Apply functions in the same way were shuffled both for S functions and Apply functions in the same way to hide
to hide the meaning of *f* and *r* prefixes. the meaning of `f` and `r` prefixes.
### Identification spoiler ### Identification spoiler:
The alternative build path `make identify` output commemorates a [previous IOCCC winner][9] which uses The alternative build path `make identify` output commemorates a [previous IOCCC
similar methods to mess with the size tool. winner][9] which uses similar methods to mess with the size tool.
### Judges remarks spoilers ### Judges' remarks spoilers:
It is not a bug, it is a feature. :) It is not a bug, it is a feature. :)
The crash is caused by a stack overflow. Conventional Unlabmda interpreters implement The crash is caused by a stack overflow. Conventional Unlabmda interpreters
the Unlambda execution stack using data structures on the heap. This interpreter uses implement the Unlambda execution stack using data structures on the heap. This
the C call stack for this purpose. In a conventional interpreter, the judges program would interpreter uses the C call stack for this purpose. In a conventional
consume memory until the user (or the operating system) would lose patience. The C call stack interpreter, the judges' program would consume memory until the user (or the
is usually of a limited size and protected, so this interpreter crashes, when the judges operating system) would lose patience. The C call stack is usually of a limited
program depletes the stack. You can always increase the maximal stack size through size and protected, so this interpreter crashes, when the judges' program
the system settings (on Linux) or compiler options (on Windows) if your Unlambda program depletes the stack. You can always increase the maximal stack size through the
needs it. Nevertheless, it will not help the judges program which quickly depletes any limit system settings (on Linux) or compiler options (on Windows) if your Unlambda
you define. program needs it. Nevertheless, it will not help the judges' program which
quickly depletes any limit you define.
The judges, pretending not to know what they are doing, created a program that grows The judges, pretending not to know what they are doing, created a program that
Unlambda execution stack indefinitely. You may consider it an Unlambda antipattern. grows Unlambda execution stack indefinitely. You may consider it an Unlambda
Normally, infinite cycles in Unlambda work by creating two functions and applying one anti-pattern. Normally, infinite cycles in Unlambda work by creating two
to the other which creates further functions to apply. The application of the function functions and applying one to the other which creates further functions to
frees its position on the stack. This program builds one ever growing function that will apply. The application of the function frees its position on the stack. This
never be complete and able to be applied. program builds one ever growing function that will never be complete and able to
be applied.
[1]: https://github.com/irori/advent-unlambda "Adventure" [1]: https://github.com/irori/advent-unlambda "Adventure"
[2]: http://www.madore.org/~david/programs/unlambda/ "Unlambda programming language" [2]: http://www.madore.org/~david/programs/unlambda/ "Unlambda programming language"