Fix #185: align problem 213 checker with the documented output format

[whuang369]

Fixes #185.

Problem

The statement for problem 213 specifies three integers per operation (l r direction), and the sample output plus every answer file in testdata/ use that format. But algorithmic/problems/213/chk.cc read only two integers per operation:

x=ouf.readInt(); y=x+m-1; z=ouf.readInt();

It derived the right endpoint from the segment length and then consumed r as the direction. For a statement-compliant l r d, the direction check (z!=0 && z!=1) fails on the first ordinary operation, so the checker quit with wrong answer Your answer is not legal before validating the sequence or computing a score. No submission following the published specification could score.

This was verified against the repository's own data: compiling the pre-fix checker and feeding it testdata/{1,2,3}.ans as the contestant output rejected all three.

Changes

• chk.cc — read l, r and d as three separate integers. The existing interval-length (r - l + 1 == x), bounds and direction validation is unchanged, so it now actually enforces the documented contract.
• testdata/1.ans — was 3 / 1 / 1 2: segment length 3 for an input with n = 2, and an operation with no direction field. Replaced with the valid 2 / 1 / 1 2 0.
• statement.txt — the prose claimed x and m share the first line while the sample printed them on separate lines. Aligned the prose to the sample, since the sample and all three answer files already agreed. Also spelled out the r - l + 1 = x and 1 <= l <= r <= n constraints that the checker enforces.
• config.yaml — time limit 2s -> the documented 5s.

Scoring logic, subtasks and the 23n / 230n thresholds are untouched.

Verification

Checker built against algorithmic/judge/include/testlib.h.

Repository answer files, before vs. after:

Case	Before	After
1.ans	wrong answer Your answer is not legal	1.0 (1 op)
2.ans	wrong answer Your answer is not legal	0.9577 (31754 ops)
3.ans	wrong answer Your answer is not legal	0.9952 (23987 ops)

The statement's own worked example (n = 5, 4 2 3 5 1) now scores 1.0.

Illegal outputs are still rejected, confirming the validation was not loosened:

• interval length != x -> Your answer is not legal
• r > n -> Your answer is not legal
• direction not in {0, 1} -> Your answer is not legal
• valid ops that do not sort the permutation -> Your answer is wrong

End-to-end, the real submission at algorithmic/solutions/213/gpt5.cpp went from rejected on all three cases to scoring 1.0 / 0.6299 / 0.6305. Every model solution in algorithmic/solutions/213/ emits the three-integer l r d form, which independently corroborates that the statement, not the checker, held the correct specification.

🤖 Generated with Claude Code