Installation and Configuration

Check the repository. The library is available on PyPI, so you can install it using pip: pip install fuzzy-dl-owl2.

Examples of supported Fuzzy Description Logic Constructs

Python Class

Description

AtomicConcept

Define an atomic concept

ChoquetIntegral

Define a choquet integral concept

ApproximationConcept

Define a tight/upper/* lower/upper approximation concept

Configuration of the MILP solver

For the configuration, create a CONFIG.ini file in the same directory used for the execution of the library. Example of your execution directory:

your_directory
├── CONFIG.ini
├── your_file.py

The file CONFIG.ini is structured as follows:

[DEFAULT]
debugPrint = False
epsilon = 0.001
maxIndividuals = -1
owlAnnotationLabel = fuzzyLabel
milpProvider = mip

Configuration Variable

Description

debugPrint

Enable/disable debugging

epsilon

Define the precision of the solution. For instance, epsilon = 0.001 means that the solution will be calculated with an accuracy to the third decimal place

maxIndividuals

Define the maximal number of individuals to handle. The value \(-1\) indicates that there is no maximum

owlAnnotationLabel

Define the Annotation label used to build the Fuzzy OWL 2 RDF/XML ontology

milpProvider

Define the MILP provider used by the reasoner. The supported providers are listed below.

Supported MILP Providers:

Provider

milpProvider

Gurobi

gurobi

CPLEX

pulp_cplex

CBC

pulp

GLPK

pulp_glpk

HiGHS

pulp_highs

MIP

mip

MILP Provider Usage and Configuration

GUROBI

  • Install gurobipy: pip install gurobipy==12.0.0

  • Download the GUROBI license from their website.

  • Add Gurobi to the PATH

MIP

  • Install Python MIP: pip install mip==1.16rc0

GLPK

  • Install GLPK v5.0 and GMP v6.3.0

  • Install Python pulp: pip install pulp==3.2.1

  • Add GLPK to the PATH

CBC

  • Install CBC

  • Install Python pulp: pip install pulp==3.2.1

  • Add CBC to the PATH

CPLEX

  • Install CPLEX v22.11

  • Install Python pulp: pip install pulp==3.2.1

  • Add CPLEX to the PATH

HiGHS

  • Install HiGHS v1.10.0

  • Install python pulp: pip install pulp==3.2.1

  • Add HiGHS to the PATH

Optional: Compile the Fast Parser

The legacy DLParser class (module fuzzy_dl_owl2.fuzzydl.parser.dl_parser) uses pyparsing to interpret FuzzyDL source files. For large knowledge bases this can dominate end-to-end runtime. A drop-in alternative is provided as DLParserFast (module fuzzy_dl_owl2.fuzzydl.parser.dl_parser_fast), which uses a hand-rolled tokenizer plus a deterministic single-pass recursive-descent parser. The grammar accepted is identical and every semantic action of the original parser is reused unchanged, so the resulting KnowledgeBase and query list are byte-equivalent.

DLParserFast is already 5-10x faster than DLParser when imported as pure Python (no backtracking, no packrat cache, no bounded-recursion bookkeeping). When the native extensions are compiled, tokenization runs in a C scanner and the hot-path modules are Cythonized, adding a further 2-3x speedup on the tokenize and dispatch hot path.

Build dependencies

Compilation is driven by build.py at the repository root, declared in pyproject.toml as a Poetry build hook ([tool.poetry.build] script = "build.py"). The build-system requirements are listed there and are installed automatically when the package is built:

cython>=3.0
setuptools>=70.0
cffi>=1.16

In addition you need:

  • A working C compiler (Xcode Command Line Tools on macOS, build-essential on Debian/Ubuntu, MSVC Build Tools on Windows).

  • A lexer generator on PATH to build the C tokenizer (see below). If none is found, build.py skips the C scanner and the parser falls back to its pure-Python tokenizer.

Lexer backend: re2c or flex

The C tokenizer is generated from a single master table (generate.py) into one of two equivalent scanner backends:

  • re2c → lexer_re2c.re (preferred).

  • flex → lexer_flex.l (fallback).

At build time build.py first tries re2c; if it is not on PATH, it tries flex. The corresponding scanner source (lexer_re2c.re or lexer_flex.l) is (re)generated by build.py automatically from generate.py, so only the tool itself needs to be installed. Both backends produce the same token stream, so install whichever is available:

# macOS (Homebrew)
brew install re2c        # or: brew install flex

# Debian/Ubuntu
sudo apt install re2c    # or: sudo apt install flex

On Windows, install one of the generators and make sure it is on PATH:

# re2c
choco install re2c       # Chocolatey
scoop install re2c       # or Scoop

# flex (win_flex.exe — expose it as `flex` on PATH)
choco install winflexbison3

On Windows the flex distribution ships the executable as win_flex.exe. Add a flex alias (or copy/rename it to flex.exe) on PATH so build.py can invoke it. Preferring re2c avoids this extra step.

Building the native extensions

The extensions are compiled in place whenever the package is built from source:

pip install .
# or
poetry build

build.py cythonizes the hot-path modules and, when re2c or flex is available, builds the C tokenizer extensions (_fdl_lexer via CFFI and _fdl_tuples via Cython) inside fuzzy_dl_owl2/fuzzydl/parser/tokenizer/. The resulting .so (or .pyd on Windows) files sit next to the Python sources; Python automatically prefers a compiled module when both files are present, so no further configuration is required.

Using DLParserFast

DLParserFast exposes the same public API as DLParser (get_kb, main, parse_string, parse_string_opt, load_config). Swap the import to opt in:

from fuzzy_dl_owl2.fuzzydl.parser import DLParserFast as DLParser

DLParser.main("example.fdl")

If the compilation step is skipped, importing DLParserFast still works and uses the pure-Python implementation.