Didactic Substeps Development

Didactic Substeps Development

This skill documents how to add pedagogical substeps to ExpliCAS engine rules so users can see concrete intermediate expressions in the step-by-step view.

Architecture Overview

There are two independent SubStep systems that both feed into the JSON output:

graph TD
  Rule["Rule (apply)"] --> Rewrite["Rewrite"]
  Rewrite -->|"poly_proof / substeps"| Step["Step (step.rs)"]
  Step -->|"Post-processing"| DidacticEnrich["didactic::enrich_steps()"]
  DidacticEnrich --> EnrichedStep["EnrichedStep { base_step, sub_steps }"]
  EnrichedStep -->|"Serialization"| JSON["eval_json.rs → SubStepJson"]

1. Engine-Level SubSteps (step.rs::SubStep)

Lightweight metadata attached to a Rewrite during rule application via Rewrite::substep(). These are title + bullet lines format (no LaTeX). They are always computed but are cheap.

// In a rule's apply() method:
Rewrite::new(before, after, "Rule description")
    .substep(step::SubStep::new(
        "Pattern Recognition",
        vec!["Detected a² - b²".into(), "a = x, b = 3".into()],
    ))

2. Didactic-Layer SubSteps (didactic::SubStep)

Rich post-processing substeps computed by didactic::enrich_steps(). These have before/after expressions with optional LaTeX. They are the ones shown in the web REPL with MathJax rendering.

// In a generate_*_substeps() function:
SubStep {
    description: "Expand products".to_string(),
    before_expr: "(t+1)(t+2)".to_string(),
    after_expr: "t² + 3t + 2".to_string(),
    before_latex: Some("(t+1)(t+2)".to_string()),
    after_latex: Some("{t}^{2} + 3t + 2".to_string()),
}

Performance Contract

[!IMPORTANT] Didactic computations MUST NOT run during normal simplification. They should only execute when step-by-step display is active.

Pattern: Gate Expensive Computation Behind Proof Data

Expensive didactic work (e.g., computing expanded polynomial forms) should be done in the rule itself only when proof data is already being computed. The rule already produces proof data as part of its correctness verification, so the marginal cost is low.

// In a rule like try_opaque_zero():
// The polynomial expansion is already computed for verification.
// We just need to also store it for display:
let expanded_form_expr =
    crate::multipoly_display::expand_additive_terms(ctx, display_expr, &display_vars);

let proof = PolynomialProofData {
    // ... standard fields ...
    expanded_form_expr,  // ← Piggyback on existing computation
};

Pattern: Post-Processing Only When Steps Are Requested

The enrich_steps() function is only called by the JSON/CLI presentation layer when steps_mode == "on". The StepMeta field in Step is Option<Box<StepMeta>> — when steps are not needed, it stays None and no allocation occurs.

Pattern: Use new_compact for Non-Display Steps

When step recording is in compact mode, use Step::new_compact() which skips the format!() call and sets meta: None.

How to Add Substeps for a New Rule

Step 1: Identify the Proof/Metadata Data

Determine what intermediate data the rule computes that would be pedagogically valuable. Common examples:

Step 2: Attach Proof Data to the Rewrite

Store the intermediate data in the Rewrite so it flows to the Step:

// Option A: Use existing poly_proof mechanism
let rewrite = Rewrite::new(before, after, "description")
    .poly_proof(proof_data);

// Option B: Use engine-level substeps (for simple title+lines)
let rewrite = Rewrite::new(before, after, "description")
    .substep(step::SubStep::new("Title", vec!["line1".into()]));

Step 3: Create a generate_*_substeps() Function

Add a new function in either didactic/nested_fractions.rs or didactic/fraction_steps.rs:

pub fn generate_my_rule_substeps(ctx: &Context, step: &Step) -> Vec<SubStep> {
    let mut sub_steps = Vec::new();
    
    // Extract proof data from the step
    let proof = match step.poly_proof() {
        Some(p) => p,
        None => return sub_steps,
    };
    
    // Build substeps with CONCRETE expressions
    if let Some(expanded_id) = proof.expanded_form_expr {
        let expanded_latex = cas_ast::LaTeXExpr {
            context: ctx,
            id: expanded_id,
        }.to_latex();
        
        sub_steps.push(SubStep {
            description: "Expand products".to_string(),
            before_expr: format!("{}", DisplayExpr { context: ctx, id: original_id }),
            after_expr: format!("{}", DisplayExpr { context: ctx, id: expanded_id }),
            before_latex: Some(original_latex),
            after_latex: Some(expanded_latex),
        });
    }
    
    sub_steps
}

Step 4: Register in the Dispatcher

In didactic/mod.rs::enrich_steps(), add a condition to call your new function:

// In the step loop inside enrich_steps():
if step.rule_name.contains("MyNewRule") {
    sub_steps.extend(generate_my_rule_substeps(ctx, step));
}

Step 5: Gate by is_chained()

If the rule uses ChainedRewrite (multi-step decomposition already visible as separate engine Steps), skip substep generation to avoid duplication:

if step.rule_name.contains("GCD") && !step.is_chained() {
    sub_steps.extend(generate_gcd_substeps(ctx, step));
}

LaTeX Rendering Contract

[!WARNING] Plain text in substeps is automatically wrapped in \text{} by the JSON layer if before_latex/after_latex are None. If you want MathJax to render math formulas, you MUST set the LaTeX fields explicitly.

Converting ExprId to LaTeX

let latex = cas_ast::LaTeXExpr {
    context: ctx,
    id: expr_id,
}.to_latex();

Converting ExprId to Plain Text

let plain = format!("{}", cas_ast::DisplayExpr {
    context: ctx,
    id: expr_id,
});

Reference: Existing Substep Generators

Concrete Example: Opaque Polynomial Identity

For (sin(u)+1)·(sin(u)+2) - sin(u)² - 3·sin(u) - 2:

Step 1 (in rule try_opaque_zero): Compute expanded_form_expr using expand_additive_terms() which splits into signed terms, expands each through multipoly independently, and recombines — gives t² + 3t + 2 - t² - 3t - 2.

Step 2 (in generate_polynomial_identity_substeps): Build three substeps:

1. "Sustitución para simplificar": Show original → Sea t = sin(u)
2. "Expresión sustituida": Show (t+1)(t+2) - t² - 3t - 2 → t² + 3t + 2 - t² - 3t - 2
3. "Todos los términos se cancelan": Show "Expandir y agrupar" → = 0

Each substep has both before_latex and after_latex so MathJax renders the math correctly.

Key Files

• cas_engine/src/didactic/mod.rs — Entry point, dispatcher, EnrichedStep/SubStep types
• cas_engine/src/didactic/nested_fractions.rs — Polynomial identity, rationalization, nested fraction generators
• cas_engine/src/didactic/fraction_steps.rs — Fraction sum, GCD factorization generators
• cas_engine/src/step.rs — Step, StepMeta, engine-level SubStep
• cas_engine/src/rule.rs — Rewrite::poly_proof(), Rewrite::substep()
• cas_engine/src/multipoly_display.rs — PolynomialProofData, expand_additive_terms()
• cas_cli/src/commands/eval_json.rs — JSON serialization, \text{} wrapping