eyeling 1.13.3 → 1.13.4

package/HANDBOOK.md

@@ -614,6 +614,8 @@ repeat
 until not changed
 ```
 
+Top-level input triples are kept as parsed (including non-ground triples such as ?X :p :o.). Groundness is enforced when adding derived facts during forward chaining, and when selecting printed/query output triples.
+
 ### 9.2 Strict-ground head optimization
 
 There is a nice micro-compiler optimization in `runFixpoint()`:

package/examples/allen-interval-calculus.n3

@@ -0,0 +1,180 @@
+# ======================================================================
+# Allen Interval Calculus
+#
+# This file illustrates the full set of 13 base relations of Allen's
+# interval algebra, encoded as forward-chaining N3 rules.
+#
+# Data model:
+#   - Any resource with :start and :end denotes a closed-open interval
+#     [start, end) over time.
+#   - :start and :end are xsd:dateTime typed literals.
+#   - Optionally, an interval may have :duration (xsd:duration).
+#
+# Built-ins (EYE / Eyeling style):
+#   - math:lessThan and math:equalTo can compare xsd:dateTime (via epoch
+#     seconds) and xsd:duration (treated as seconds).
+#   - Timestamp arithmetic uses:
+#       (dateTime duration) math:sum dateTime
+#       (dateTime1 dateTime2) math:difference duration
+#       (dateTime duration) math:difference dateTime
+#     where xsd:duration uses a simplified seconds model.
+# ======================================================================
+
+@prefix :     <http://example.org/allen#>.
+@prefix math: <http://www.w3.org/2000/10/swap/math#>.
+@prefix xsd:  <http://www.w3.org/2001/XMLSchema#>.
+
+# -------------------------------------------------------------
+# Optional interval completion rules (derive missing endpoints)
+# -------------------------------------------------------------
+
+# If start + duration given, derive end.
+{
+  ?I :start ?s; :duration ?d.
+  ( ?s ?d ) math:sum ?e.
+} => { ?I :end ?e. }.
+
+# If end + duration given, derive start.
+{
+  ?I :end ?e; :duration ?d.
+  ( ?e ?d ) math:difference ?s.
+} => { ?I :start ?s. }.
+
+# If start + end given, derive duration (seconds-only lexical output).
+{
+  ?I :start ?s; :end ?e.
+  ( ?e ?s ) math:difference ?d.
+} => { ?I :duration ?d. }.
+
+# ----------------------------------------------------
+# 7 "forward" base relations (the rest are converses)
+# ----------------------------------------------------
+
+# before(I,J) <=> end(I) < start(J)
+{
+  ?I :end ?eI.
+  ?J :start ?sJ.
+  ?eI math:lessThan ?sJ.
+} => { ?I :before ?J. }.
+
+# meets(I,J) <=> end(I) = start(J)
+{
+  ?I :end ?eI.
+  ?J :start ?sJ.
+  ?eI math:equalTo ?sJ.
+} => { ?I :meets ?J. }.
+
+# overlaps(I,J) <=> start(I) < start(J) < end(I) < end(J)
+{
+  ?I :start ?sI; :end ?eI.
+  ?J :start ?sJ; :end ?eJ.
+  ?sI math:lessThan ?sJ.
+  ?sJ math:lessThan ?eI.
+  ?eI math:lessThan ?eJ.
+} => { ?I :overlaps ?J. }.
+
+# starts(I,J) <=> start(I)=start(J) AND end(I) < end(J)
+{
+  ?I :start ?s;  :end ?eI.
+  ?J :start ?s;  :end ?eJ.
+  ?eI math:lessThan ?eJ.
+} => { ?I :starts ?J. }.
+
+# during(I,J) <=> start(J) < start(I) AND end(I) < end(J)
+{
+  ?I :start ?sI; :end ?eI.
+  ?J :start ?sJ; :end ?eJ.
+  ?sJ math:lessThan ?sI.
+  ?eI math:lessThan ?eJ.
+} => { ?I :during ?J. }.
+
+# finishes(I,J) <=> end(I)=end(J) AND start(J) < start(I)
+{
+  ?I :start ?sI; :end ?e.
+  ?J :start ?sJ; :end ?e.
+  ?sJ math:lessThan ?sI.
+} => { ?I :finishes ?J. }.
+
+# equals(I,J) <=> start(I)=start(J) AND end(I)=end(J)
+{
+  ?I :start ?s; :end ?e.
+  ?J :start ?s; :end ?e.
+} => { ?I :equals ?J. }.
+
+# ------------------------------------------------------
+# 6 converse relations to complete the 13 base relations
+# ------------------------------------------------------
+
+{ ?I :before ?J. }    => { ?J :after ?I. }.
+{ ?I :meets ?J. }     => { ?J :metBy ?I. }.
+{ ?I :overlaps ?J. }  => { ?J :overlappedBy ?I. }.
+{ ?I :starts ?J. }    => { ?J :startedBy ?I. }.
+{ ?I :during ?J. }    => { ?J :contains ?I. }.
+{ ?I :finishes ?J. }  => { ?J :finishedBy ?I. }.
+
+# -------------------------------------------------------
+# Sanity check: flag invalid intervals where start >= end
+# -------------------------------------------------------
+
+{
+  ?I :start ?s; :end ?e.
+  ?e math:lessThan ?s.
+} => { ?I :invalidInterval true. }.
+
+{
+  ?I :start ?s; :end ?e.
+  ?s math:equalTo ?e.
+} => { ?I :invalidInterval true. }.
+
+# ------------------------------------------
+# Example intervals (xsd:dateTime endpoints)
+# ------------------------------------------
+
+# A canonical set to demonstrate multiple relations
+:A :start "2026-02-18T10:00:00Z"^^xsd:dateTime;
+   :end   "2026-02-18T12:00:00Z"^^xsd:dateTime.
+
+:B :start "2026-02-18T13:00:00Z"^^xsd:dateTime;
+   :end   "2026-02-18T15:00:00Z"^^xsd:dateTime.
+
+# :A meets :C (A ends at 12:00; C starts at 12:00)
+:C :start "2026-02-18T12:00:00Z"^^xsd:dateTime;
+   :end   "2026-02-18T14:00:00Z"^^xsd:dateTime.
+
+# :A overlaps :D (10:00 < 11:00 < 12:00 < 13:00)
+# and :D meets :B (D ends at 13:00; B starts at 13:00)
+:D :start "2026-02-18T11:00:00Z"^^xsd:dateTime;
+   :end   "2026-02-18T13:00:00Z"^^xsd:dateTime.
+
+# :A equals :E
+:E :start "2026-02-18T10:00:00Z"^^xsd:dateTime;
+   :end   "2026-02-18T12:00:00Z"^^xsd:dateTime.
+
+# :F starts :A
+:F :start "2026-02-18T10:00:00Z"^^xsd:dateTime;
+   :end   "2026-02-18T11:00:00Z"^^xsd:dateTime.
+
+# :G finishes :A
+:G :start "2026-02-18T11:00:00Z"^^xsd:dateTime;
+   :end   "2026-02-18T12:00:00Z"^^xsd:dateTime.
+
+# :A during :H (and thus :H contains :A)
+:H :start "2026-02-18T09:00:00Z"^^xsd:dateTime;
+   :end   "2026-02-18T16:00:00Z"^^xsd:dateTime.
+
+# -----------------------------------------------
+# Examples using xsd:duration + derived endpoints
+# -----------------------------------------------
+
+# :I is given as start + duration; :end will be derived via math:sum
+:I :start    "2026-02-18T16:00:00Z"^^xsd:dateTime;
+   :duration "PT2H"^^xsd:duration.
+
+# :J meets :I once :I's end is derived (J ends at 16:00; I starts at 16:00)
+:J :start "2026-02-18T15:00:00Z"^^xsd:dateTime;
+   :end   "2026-02-18T16:00:00Z"^^xsd:dateTime.
+
+# :K is given as start + duration; it will derive end=14:00 and thus finishes :C
+:K :start    "2026-02-18T13:30:00Z"^^xsd:dateTime;
+   :duration "PT30M"^^xsd:duration.
+

package/examples/deck/odrl-dpv-risk-ranked.md

@@ -0,0 +1,251 @@
+# ODRL + DPV Risk Assessment
+
+## Ranked, explainable output from machine-readable “Terms of Service”
+
+This deck explains how an agreement is modeled in **ODRL**, how risks are expressed in **DPV**, and how **N3 rules** connect the two into a ranked report. ([Playground][1])
+
+---
+
+## The idea
+
+We want ToS / policy clauses that are:
+
+- **Readable by humans** (the actual clause text)
+- **Processable by machines** (permissions, prohibitions, duties, constraints)
+- **Auditable** (why a risk was flagged)
+- **Actionable** (what mitigations to add)
+- **Prioritized** (ranked by score)
+
+This example does that by combining **ODRL** (policy structure) + **DPV** (risk vocabulary) + **N3 rules** (logic). ([Playground][1])
+
+---
+
+## Why ODRL matters here
+
+ODRL is used to encode the _normative_ structure of agreements:
+
+- **Permission**: something is allowed
+- **Prohibition**: something is disallowed
+- **Duty**: something must be done (e.g., inform)
+- **Constraint**: conditions like “noticeDays ≥ 14”
+
+This turns ToS clauses into a structured “policy graph” you can reason over. ([ODRL Vocabulary & Expression 2.2][2])
+
+---
+
+## Why DPV matters here
+
+DPV provides shared terms to describe privacy-related concepts, including:
+
+- **dpv:Risk**
+- consequences / impacts
+- severity & risk level (via the DPV Risk extension)
+
+So the output isn’t just “something seems bad”, but _typed, interoperable risks_ that other systems can understand. ([DPV Risk & Impact Assessment][3])
+
+---
+
+## What the file contains (5 parts)
+
+1. **Consumer profile** (needs + importance weights)
+2. **Agreement** as an **ODRL policy graph** + linked clause text
+3. **Risk rules**: patterns over ODRL → create DPV risks + mitigations
+4. **Score + severity/level** classification
+5. **Ranked explainable output** strings
+
+All in one Notation3 (N3) program. ([Playground][1])
+
+---
+
+## Part 1 — Consumer profile (what the user cares about)
+
+The example profile defines four “needs”, each with an importance weight:
+
+| Need                       | Meaning                                | Importance |
+| -------------------------- | -------------------------------------- | ---------: |
+| Data cannot be removed     | provider shouldn’t remove account/data |         20 |
+| Changes need notice        | must notify ≥ 14 days                  |         15 |
+| No sharing without consent | explicit consent required              |         12 |
+| Data portability           | must allow export                      |         10 |
+
+These weights later boost the risk score when a need is violated. ([Playground][1])
+
+---
+
+## Part 2 — Agreement modeled as ODRL
+
+Inside a quoted graph (`:policyGraph { ... }`) the policy defines:
+
+- **C1** Permission to remove account/data
+- **C2** Permission to change terms with an **inform duty** and **noticeDays ≥ 3**
+- **C3** Permission to share user data (no consent safeguard)
+- **C4** Prohibition to export data (blocks portability)
+
+Each ODRL rule links to a `:Clause` resource that stores the human text. ([Playground][1])
+
+---
+
+## ODRL clause pattern (how to read it)
+
+A typical ODRL rule here looks like:
+
+- **assigner**: provider
+- **assignee**: consumer
+- **action**: (removeAccount / shareData / changeTerms / exportData)
+- **target**: (UserAccount / UserData / AgreementText)
+- optional **duty** (e.g., inform)
+- optional **constraint** (e.g., noticeDays threshold)
+
+That structure is what the logic rules match on. ([Playground][1])
+
+---
+
+## Part 3 — The logic bridge: N3 rules
+
+Each risk rule follows the same recipe:
+
+1. **Match** a clause in the policy graph (`log:includes`)
+2. **Detect missing safeguards** (`log:notIncludes`) or insufficient safeguards (comparisons)
+3. **Create** a DPV risk instance (`dpv:Risk`) + risk source + explanation text
+4. **Attach mitigations** as `dpv:RiskMitigationMeasure`
+5. **Store** a numeric score seed (`:scoreRaw`)
+
+Key N3 tools you’ll see:
+
+- `log:includes` / `log:notIncludes` for scoped graph checks ([Notation3 Language][4])
+- `log:skolem` to mint stable identifiers for risks/measures ([Playground][1])
+- `string:format`, `math:sum`, `math:difference`, comparisons, etc. ([Playground][1])
+
+---
+
+## Deep dive: Rule R3 (share data without consent)
+
+**Natural language translation:**
+
+> If the agreement permits sharing user data, and the consumer requires “no sharing without explicit consent”, and the policy graph does **not** contain a consent constraint for that sharing permission, then generate a DPV risk “unwanted disclosure”, explain it, score it, and suggest adding a consent constraint.
+
+This is exactly what `log:includes` + `log:notIncludes` is doing. ([Playground][1])
+
+---
+
+## R3 “missing safeguard” pattern (conceptual)
+
+```n3
+?G log:includes    { :PermShareData odrl:action tosl:shareData . } .
+?G log:notIncludes { :PermShareData odrl:constraint [
+    odrl:leftOperand tosl:consent ;
+    odrl:operator odrl:eq ;
+    odrl:rightOperand true
+] . } .
+```
+
+Result: create `dpv:Risk` + add mitigation “Add explicit consent constraint before data sharing.” ([Playground][1])
+
+---
+
+## Part 4 — Scoring and DPV risk levels
+
+### Scoring (simple and explainable)
+
+Each rule computes:
+
+- `:scoreRaw = base + needImportance`
+- then caps at **100**
+
+### Mapping score → severity/level
+
+- **80–100** → High severity / High risk
+- **50–79** → Moderate
+- **0–49** → Low
+
+This gives a consistent DPV-style classification (`dpv:hasSeverity`, `dpv:hasRiskLevel`). ([Playground][1])
+
+---
+
+## What the example will rank (with the given numbers)
+
+From the file’s constants + importance weights:
+
+1. **C1 account removal w/o notice + inform** base 90 + 20 = 110 → capped **100** (High)
+2. **C3 sharing w/o consent** base 85 + 12 = **97** (High)
+3. **C2 terms change notice too short (3 < 14)** base 70 + 15 = **85** (High)
+4. **C4 export prohibited (no portability)** base 60 + 10 = **70** (Moderate)
+
+So the “worst” risks appear first. ([Playground][1])
+
+---
+
+## Part 5 — Ranked, explainable output
+
+Instead of “printing during reasoning”, the program emits facts like:
+
+- `log:outputString "..."`
+
+Then Eyeling’s `--strings` / `-r` mode collects and sorts them deterministically. ([Handbook Inside Eyeling][5])
+
+To force ranking, it uses an **inverse score key**:
+
+- `inv = 1000 - score`
+- smaller `inv` → higher score → printed first
+
+That’s why high-risk items appear at the top. ([Playground][1])
+
+---
+
+## What makes it “explainable”
+
+Every risk carries:
+
+- **Which clause** it came from (`:aboutClause`, clauseId + text)
+- **Which need** it violated (`:violatesNeed`)
+- A human explanation string (`dct:description`, built with `string:format`)
+- Suggested **mitigations**, each with a description and even a “patch-like” triple snippet (`:suggestAdd { ... }`)
+
+So you can show a ranked report _and_ justify every item. ([Playground][1])
+
+---
+
+## Why this ODRL + DPV combo is powerful
+
+- **ODRL** gives you the “contract logic” backbone (may/must/must-not + conditions)
+- **DPV** gives you the “privacy/risk language” that tools can share
+- **N3** glues them with rules that are:
+  - easy to audit
+  - easy to extend
+  - deterministic to run
+
+This is a practical path from “legal-ish text” → “structured policy” → “ranked risk insights”.
+
+---
+
+## How you’d extend this in real life
+
+1. **Add more needs** (e.g., retention limits, security measures, breach notice)
+2. **Model more clause types** in ODRL (more actions, constraints, duties)
+3. **Write additional risk rules**, each with:
+   - pattern match
+   - missing/weak safeguard test
+   - DPV risk type + mitigation
+
+4. Tune scoring:
+   - different bases per risk category
+   - incorporate likelihood, data sensitivity, etc.
+
+This stays explainable because it remains rule-based. ([Playground][1])
+
+---
+
+## Closing takeaway
+
+This file is a compact demo of:
+
+- ODRL as **machine-readable agreement structure**
+- DPV as **machine-readable privacy risk output**
+- N3 reasoning as the **transparent logic** connecting them
+- A ranked report that’s **deterministic** and **explainable**
+
+[1]: https://eyereasoner.github.io/eyeling/demo?url=https://raw.githubusercontent.com/eyereasoner/eyeling/refs/heads/main/examples/odrl-dpv-risk-ranked.n3 'Playground'
+[2]: https://www.w3.org/TR/odrl-vocab/ 'ODRL Vocabulary & Expression 2.2'
+[3]: https://dev.dpvcg.org/dpv/modules/risk 'Risk and Impact Assessment'
+[4]: https://w3c.github.io/N3/spec/ 'Notation3 Language'
+[5]: https://eyereasoner.github.io/eyeling/HANDBOOK 'Handbook Inside Eyeling'