perf: Replace ISO date parser's split/collect with fixed-position structural checks

[Copilot]

The ISO format parser (YYYY-MM, YYYY-MM-DD) was using split().collect::<Vec<_>>(), allocating a Vec on every successful parse — inconsistent with the already-optimized slash-format parser.

Changes

• New parse_iso_date dispatcher: selects between two dedicated functions based on string length (7 or 10); emits a "Too many - separators" error for lengths with excess hyphens
• New parse_iso_7: handles YYYY-MM (exactly 7 bytes) — validates digits at fixed positions 0–3 and 5–6, and a - at position 4, using a single compound boolean; no loop, no allocation
• New parse_iso_10: handles YYYY-MM-DD (exactly 10 bytes) — validates digits at fixed positions 0–3, 5–6, and 8–9, and - separators at positions 4 and 7, using a single compound boolean; no loop, no allocation
• Removed parse_iso_month_year and parse_iso_full_date: consolidated into the two new functions

// Before — allocates Vec on every successful parse
let parts: Vec<&str> = trimmed.split(DATE_SEPARATOR).map(str::trim).collect();
match parts.len() { ... }

// After — length dispatch into fixed-position structural checks; zero allocation
Self::parse_iso_date(trimmed)

---

🔒 GitHub Advanced Security automatically protects Copilot coding agent pull requests. You can protect all pull requests by enabling Advanced Security for your repositories. Learn more about Advanced Security.

[semanticdiff-com]

Review changes with  

Changed Files

File	Status
src/lib.rs	33% smaller

[tcrypt25519]

@copilot There's no reason to bridge like 20 times for this. You can get it down to 2 if you do this with separate fns for 8 and 10.

Structural ISO 8601 date validation algorithm

Goal: validate only the character structure of a string as either:

YYYY-MM (length 8)

YYYY-MM-DD (length 10)

No semantic checks. Only shape.

---

Step-by-step algorithm

1. Read input

Let s be the character array.

Let len be its length.

---

2. Length guard

if len != 8 AND len != 10:
return false

---

3. Initialize counters

digitSum = 0
hyphenSum = 0

---

4. Validate required digit positions

Always check:

indices: 0,1,2,3,5,6

For each index i above:

if s[i] is digit:
digitSum += 1

If len == 10, also check:

indices: 8,9

---

5. Validate hyphen positions

Always check:

if s[4] == '-':
hyphenSum += 1

If len == 10:

if s[7] == '-':
hyphenSum += 1

---

6. Final structural verdict

if len == 8:
return digitSum == 6 AND hyphenSum == 1

if len == 10:
return digitSum == 8 AND hyphenSum == 2

---

7. End

Result is true only if the string matches the required structural layout exactly.

[Copilot]

@copilot There's no reason to bridge like 20 times for this. You can get it down to 2 if you do this with separate fns for 8 and 10.

Structural ISO 8601 date validation algorithm...

Done in e2d4624. The byte-walker is replaced by two fixed-position functions: parse_iso_7 (YYYY-MM, 7 bytes) and parse_iso_10 (YYYY-MM-DD, 10 bytes). A thin parse_iso_date dispatcher selects between them by string length. Each inner function does a single compound boolean check on fixed indices — no loop, no allocation. Note: YYYY-MM is 7 bytes (not 8 as listed in the spec), so the dispatch key is 7 and 10.