This is a Python implementation of Douglas Hofstadter formal systems, from his book Gödel, Escher, Bach: An Eternal Golden Braid (commonly GEB).
In fact, you may define your own formal systems using a quite simple syntax, close to free text. Examples for MIU, pg, fg and NDP formal systems from the book are implemented in directory definitions.
A main Python script gives you possibilities to play with the formal system, including:
- axiom schema support (iteration, decision procedure)
- theorem step by step generation (using different algorithms)
- theorem derivation
The MIU system may be define with:
axioms:
- MI
rules:
- x is .*, xI => xIU
- x is .*, Mx => Mxx
- x is .*, y is .*, xIIIy => xUy
- x y .* , xUUy => xy
The underlying syntax is YAML (see raw format). You can define one or several axioms, or even an infinite number of axioms using a schema, as in the pg formal system:
axioms:
- x is -+, xp-gx-
rules:
- x y z are -+, xpygz => xpy-gz-
Axiom definitions should be formatted like this ([]
means this is optional):
[def_1, [def_2, ...]] expr
Where:
def_i
is an optional definition of wildcard, using a regular expression, for example:- ".*" may be anything including the empty string
- "-+" is a string composed of "-"
The definitions are written using char [is] regexp
or char1 char2 [are] regexp
if different wildcards have the same definition. Note that you should use only one character for wildcard definition.
expr
is the axiom expression
Rules for theorem production should be formatted like this:
[def_1, [def_2, ...]] cond_1 [and cond_2 [and ...]] => th_1 [and th_2 [and ...]]
Where:
def_i
is the same as beforecond_i
is a required theorem, in order to produce new theorems (separated by and if several conditions)th_i
is a produced theorems with the rule
Install with:
$ python setup.py install --user
A script should be put in ~/.local/bin
, make sure this path is in your $PATH
:
$ export PATH=$PATH:~/.local/bin
If installation is successful, run the tests with:
$ cd tests
$ python test_formalsystems.py -v
After installation, you should have the main script FormalSystemsMain.py
deployed somewhere where you $PATH
points to, under the name FormalSystems
.
If it is not the case, you can always execute the script directly, assuming the dependencies are properly installed (just pyyaml and LEPL).
Usage of the main script is fully documented in --help
argument.
You may generate theorems step by step if the number of axioms is finite:
$ FormalSystems definitions/MIU.yaml --iteration 3
> Finite number of axioms, using step algorithm
STEP 1: MI
P (1) x is .*, xI => xIU for MI gives MIU
P (2) x is .*, Mx => Mxx for MI gives MII
. (3) x is .*, y is .*, xIIIy => xUy for MI
. (4) x y .* , xUUy => xy for MI
STEP 2: MIU/MII
P (1) x is .*, xI => xIU for MII gives MIIU
. (1) x is .*, xI => xIU for MIU
P (2) x is .*, Mx => Mxx for MII gives MIIII
P (2) x is .*, Mx => Mxx for MIU gives MIUIU
. (3) x is .*, y is .*, xIIIy => xUy for MII
. (3) x is .*, y is .*, xIIIy => xUy for MIU
. (4) x y .* , xUUy => xy for MII
. (4) x y .* , xUUy => xy for MIU
STEP 3: MIIU/MIIII/MIUIU
Or using a bucket where axioms are thrown and theorems computed iteratively if the number of axioms is infinite:
$ FormalSystems definitions/pg.yaml --iteration 4
> Infinite number of axioms, using bucket algorithm
[Adding -p-g-- to bucket]
=== BUCKET 1: -p-g--
P (1) x y z are -+, xpygz => xpy-gz- for -p-g-- gives -p--g---
[Adding --p-g--- to bucket]
=== BUCKET 2: -p--g---/--p-g---
P (1) x y z are -+, xpygz => xpy-gz- for -p--g--- gives -p---g----
P (1) x y z are -+, xpygz => xpy-gz- for --p-g--- gives --p--g----
[Adding ---p-g---- to bucket]
=== BUCKET 3: -p---g----/--p--g----/---p-g----
P (1) x y z are -+, xpygz => xpy-gz- for -p---g---- gives -p----g-----
P (1) x y z are -+, xpygz => xpy-gz- for ---p-g---- gives ---p--g-----
P (1) x y z are -+, xpygz => xpy-gz- for --p--g---- gives --p---g-----
[Adding ----p-g----- to bucket]
=== BUCKET 4: -p----g-----/---p--g-----/--p---g-----/----p-g-----
Options are available to display theorem derivation as well:
$ FormalSystems definitions/NDP.yaml --quiet --derivation P-----
> Infinite number of axioms, using bucket algorithm
> Rule with several parents, using recursivity
=== BUCKET 1: --NDP-
=== BUCKET 2: --NDP---/-SD--/P--
=== BUCKET 3: --NDP-----/---SD--/---NDP--
=== BUCKET 4: --NDP-------/---NDP-----/-----SD--/P---/---NDP-
=== BUCKET 5: --NDP---------/---NDP--------/---NDP----/-------SD--/-----SD---/-SD---/----NDP---
=== BUCKET 6: ---NDP-----------/----NDP-------/---NDP-------/--NDP-----------/---------SD--/----NDP-
=== BUCKET 7: ----NDP-----------/----NDP-----/---NDP----------/---NDP--------------/--NDP-------------/-----------SD--/-------SD---/-SD----/----NDP--
=== BUCKET 8: ----NDP---------/----NDP---------------/---NDP-------------/---NDP-----------------/--NDP---------------/----NDP------/-------------SD--/-------SD----/-----SD----/-----------SD---/-----NDP-
=== BUCKET 9: --NDP-----------------/-----NDP------/----NDP-------------/---NDP--------------------/---NDP----------------/----NDP----------/----NDP-------------------/---------------SD--/-SD-----/-------------SD---/-----------SD----/P-----/-----NDP--
=== Theorem P----- found, derivation:
[1 ] Axiom gives --NDP-
[2 ] (1) x y are -+, xNDPy => xNDPxy for --NDP- gives --NDP---
[3 ] Axiom gives ---NDP--
[3 ] (1) x y are -+, xNDPy => xNDPxy for --NDP--- gives --NDP-----
[4 ] Axiom gives ----NDP-
[4 ] (1) x y are -+, xNDPy => xNDPxy for ---NDP-- gives ---NDP-----
[4 ] (2) z is -+, --NDPz => zSD-- for --NDP----- gives -----SD--
[5 ] (1) x y are -+, xNDPy => xNDPxy for ----NDP- gives ----NDP-----
[5 ] (3) x z are -+, zSDx and x-NDPz => zSDx- for -----SD-- and ---NDP----- gives -----SD---
[6 ] (3) x z are -+, zSDx and x-NDPz => zSDx- for -----SD--- and ----NDP----- gives -----SD----
[7 ] (4) z is -+, z-SDz => Pz- for -----SD---- gives P-----
Some tests using doctests:
>>> from formalsystems.formalsystems import FormalSystem, Theorem
MIU formal system:
>>> fs = FormalSystem()
>>> fs.read_formal_system('../definitions/MIU.yaml')
>>> r = fs.apply_rules_step(fs.iterate_over_schema(), step=4, verbose=False)
STEP 1: MI
STEP 2: MIU/MII
STEP 3: MIIU/MIIII/MIUIU
STEP 4: MIIIIU/MIIIIIIII/MIIUIIU/MIUIUIUIU/MIU/MUI
>>> print [str(a) for a in fs.iterate_over_schema()]
['MI']
pg formal system:
>>> fs = FormalSystem()
>>> fs.read_formal_system('../definitions/pg.yaml')
>>> r = fs.apply_rules_bucket_till(fs.iterate_over_schema(), max_turns=4, verbose=False)
=== BUCKET 1: -p-g--
=== BUCKET 2: -p--g---/--p-g---
=== BUCKET 3: -p---g----/--p--g----/---p-g----
=== BUCKET 4: -p----g-----/---p--g-----/--p---g-----/----p-g-----
>>> r = fs.apply_rules_bucket_till(fs.iterate_over_schema(), min_len=9, verbose=False)
=== BUCKET 1: -p-g--
=== BUCKET 2: -p--g---/--p-g---
=== BUCKET 3: -p---g----/--p--g----/---p-g----
NDP formal system:
>>> fs = FormalSystem()
>>> fs.read_formal_system('../definitions/NDP.yaml')
>>> r = fs.apply_rules_bucket_till(fs.iterate_over_schema(), max_turns=2, full=True, verbose=False)
=== BUCKET 1: --NDP-
=== BUCKET 2: --NDP---/-SD--/P--
Successful derivation:
>>> fs = FormalSystem()
>>> fs.read_formal_system('../definitions/NDP.yaml')
>>> r = fs.derivation_asc(fs.iterate_over_schema(), Theorem('P-----'), full=True, max_turns=10)
<BLANKLINE>
...
=== Theorem P----- found, derivation:
...
Failed derivation:
>>> fs = FormalSystem()
>>> fs.read_formal_system('../definitions/MIU.yaml')
>>> r = fs.derivation_step(fs.iterate_over_schema(), Theorem('MIUIU'), step=5)
<BLANKLINE>
...
=== Theorem MIUIU found, derivation:
...
>>> r = fs.derivation_step(fs.iterate_over_schema(), Theorem('MU'), step=5)
<BLANKLINE>
...
=== Theorem MU not found