Ferrifuzz for VS Code
Editor support for ferrifuzz, fuzzy logic for Python
with a Rust core: rule-string highlighting, completion and checks, snippets, and a check that
VS Code's Python interpreter can see ferrifuzz.
Needs the Python extension
and ferrifuzz 0.2.0 or newer in the interpreter VS Code uses (pip install ferrifuzz).
Rule highlighting
Rule strings inside Python are highlighted — keywords, input/output variables, NOT,
hedges, terms, parentheses and WITH weights each get their own color:
system += "IF temp IS NOT very hot OR hum IS high THEN fan IS slow AND alarm IS on"
system += "if (temp is hot or temp is very warm) and wind is calm then fan is fast with 0.8"
RULES = """
IF service IS poor THEN tip IS low
IF service IS excellent THEN tip IS high
"""
Only strings (or lines of a triple-quoted string) that start with IF <var> IS (any case,
optionally after () and contain THEN are touched, so ordinary strings and docstrings keep
their normal color. f-strings are left alone.
| Part |
Scope |
IF, THEN |
keyword.control.conditional.ferrifuzz |
AND, OR |
keyword.operator.logical.ferrifuzz |
IS, WITH |
keyword.operator.word.ferrifuzz |
| weight number |
constant.numeric.ferrifuzz |
(, ) |
punctuation.section.parens.ferrifuzz |
NOT |
keyword.operator.logical.negation.ferrifuzz |
| variable |
variable.other.input.ferrifuzz |
hedge (very, somewhat, ...) |
storage.modifier.hedge.ferrifuzz |
| term |
entity.name.type.term.ferrifuzz |
Recolor with editor.tokenColorCustomizations → textMateRules if your theme is too quiet.
Rule completion and checks
Inside a rule string the editor suggests what can come next, read from the ff.Variable(...)
definitions, var["term"] = ... / .automf(...) terms and add_input / add_output calls
in the same file:
| After |
Suggests |
IF / AND / OR / ( |
input variables |
THEN / AND after THEN |
output variables |
<var> |
IS |
IS / IS NOT |
that variable's terms, NOT, hedges (very, somewhat, ...) |
| a hedge |
that variable's terms |
| a term |
AND OR THEN (before THEN), AND WITH (after) |
Keywords follow the rule's own case (if → is, then, ...). Type a space to see the next
suggestions.
Rule strings that go into a system in the same file (system += "...", add_rule("..."),
add_rules([...]), or a constant later passed to those) are checked as you type:
- a term the variable does not have —
'service' has no term 'excelent'. Did you mean 'excellent'?
- a misspelled variable —
Unknown variable 'servise'. Did you mean 'service'?
- an input after THEN or an output before it
"Did you mean" warnings come with a quick fix (Cmd/Ctrl+.). The
checks stay quiet about anything the file does not pin down — variables imported from another
module, terms added in a loop, rules parsed on their own with Rule.from_string — and
suggestions use the same edit-distance limits as ferrifuzz's own error messages.
Turn them off with ferrifuzz.ruleCompletion.enabled / ferrifuzz.ruleDiagnostics.enabled.
Snippets
Type the prefix in a Python file and press Tab. Dropdowns let you pick
t-norms, defuzzifiers, type reduction and so on.
| Prefix |
Inserts |
ffimport |
import ferrifuzz as ff |
ffmamdani |
complete Mamdani system (operator / defuzz pickers) |
ffsugeno |
complete Sugeno/TSK system with a linear consequent |
ffit2sys |
complete interval type-2 system with an FOU helper and compute_interval |
ffvar |
variable with low/medium/high terms |
ffautovar |
ff.Variable(...).automf(3 \| 5 \| 7, shape=...) |
ffterm |
var["term"] = ff.mf.triangle(...) |
fftri fftrap ffgauss ffbell ffsig ffsingleton ffit2 |
membership functions |
ffrule ffrule2 ffrules ffruleobj |
one rule, two-clause rule, rule list, weighted Rule(...) |
ffcompute ffbatch ffloop |
one row, a NumPy column, a closed-loop simulation with evaluate |
ffvalidate fftry ffstrengths |
list errors/warnings, catch ModelError/InputError, print rule firing strengths |
ffhedge ffand ffor |
hedge / t-norm / t-conorm on raw degrees |
ffplotvar ffplotout ffplotcurve ffsurface ffshow |
ferrifuzz.plot: terms, output set, response curve, 3-D surface, plt.show() |
fffts |
fuzzy time series forecast + evaluate on the Alabama enrollment data |
Method-name arguments (defuzz_method=, and_method=, ...) also autocomplete through the
type stubs shipped with the ferrifuzz package.
Interpreter check
Completions and hover docs come from the type stubs inside the installed ferrifuzz
package, so they only work when the interpreter selected in VS Code has ferrifuzz 0.2.0 or
newer. When a Python file imports ferrifuzz and the selected interpreter has no ferrifuzz,
or an older one, a warning offers:
- Use .venv — when the workspace's
.venv / venv has a usable ferrifuzz, switch to it;
- Select Interpreter — open the Python extension's picker;
- Don't Show Again — stop warning about that interpreter.
The check reads package metadata only (importlib.metadata), never imports ferrifuzz, and
runs once per interpreter per session. It needs the Python extension (ms-python.python).
| Command / setting |
|
| Ferrifuzz: Check Interpreter for Completions |
check the active file's interpreter now |
| Ferrifuzz: Reset Interpreter Warnings |
undo every "Don't Show Again" |
ferrifuzz.interpreterCheck.enabled |
turn the automatic check off |
Install
Search for Ferrifuzz in the Extensions view: the VS Code Marketplace, or Open VSX in
Cursor, VSCodium and Windsurf. From a .vsix file: Extensions → ⋯ → Install from VSIX…, or
code --install-extension ferrifuzz-<version>.vsix.
Development
npm install
npm run package # → ferrifuzz-<version>.vsix
npm run build:snippets # regenerate snippets/python.code-snippets from tools/build_snippets.py
npm run test:snippets # expand every snippet with its defaults and execute it (needs ferrifuzz, numpy, matplotlib)
npm run test:grammar # tokenize tools/grammar_sample.py with VS Code's Python grammar + the injection
npm test # the three below (CI runs this)
npm run test:check # interpreter check against throwaway venvs (needs python3)
npm run test:rules # rule completion / checks, plus a no-false-alarm scan of examples/ and tests/
npm run test:extension # extension.js against a fake vscode module
Edit snippets in tools/build_snippets.py, not in the generated JSON.
Publishing
Bump version here and in CHANGELOG.md, run npm test, then:
npx vsce login ParsaBordbar2003 # once; Azure DevOps token with Marketplace → Manage scope
npm run publish:vsce # VS Code Marketplace
OVSX_PAT=... npm run publish:ovsx # Open VSX (Cursor, VSCodium, Windsurf); token from open-vsx.org
