[rayokota]

Port the excessive sequence length guardrail from the JS reference implementation: a new optional sequence option caps the length of any sequence created during evaluation, including intermediate sequences, raising D2015 when exceeded. The range operator and $append also check the size of their result up front. Also update the D2014 message to match the 1e7 limit that is already enforced.

[Copilot]

🟡 Changes recommended

Sequence-limit enforcement remains incorrect: wildcard results can bypass the limit after a bounded accumulator becomes an unbounded list, while an off-by-one range length rejects valid exact-limit results.

2 open issues

• Preserve sequence bounds when copying wildcard results · New
• Fix off-by-one range length at the sequence limit · New

What changed in this PR

Adds an optional sequence-length guardrail to the Python JSONata evaluator, extending its protections for untrusted expressions.

Changes:

• Introduces the sequence option, bounded sequence construction, and limit restoration after evaluation.
• Adds upfront range and $append checks, introduces D2015, and corrects D2014’s documented limit to 1e7.
• Documents the option and adds tests for limits, intermediate sequences, $eval, and restoration.

File	Description
tests/​string_test.py	Adds sequence guardrail regression tests.
src/​jsonata/​utils.py	Adds thread-local limits and bounded sequence lists.
src/​jsonata/​jsonata.py	Exposes the option and applies it during evaluation.
src/​jsonata/​jexception.py	Adds D2015 and corrects D2014’s message.
src/​jsonata/​functions.py	Checks $append result size before copying.
README.md	Documents the sequence-length guardrail.

_{Review effort: Balanced}

[Copilot]

🟡 Changes recommended

Sequence-limit errors report the attempted length instead of the configured maximum, and returned lists retain bounds that incorrectly restrict caller mutations after evaluation.

1 open issue

• Avoid enforcing sequence bounds after evaluation · New

1 previously missed issue

In code that hasn't changed since last review

Report the configured maximum in D2015 errors

src/​jsonata/​utils.py:129

This passes the attempted length into D2015, but its message describes that value as the configured maximum. For example, [0..1001] with sequence=1000 reports “The maximum sequence length of 1002 was exceeded,” and $append has the same problem. Pass max_length instead so the message and JException.current agree with the configured limit and with BoundedJList._check.

2 resolved since last review

• Preserve sequence bounds when copying wildcard results
• Fix off-by-one range length at the sequence limit

_{Review effort: Balanced}

[Copilot]

🔵 Needs a closer look

The sequence limit is checked only after wildcard flattening copies the entire array, so intermediate allocations can still exhaust memory.

No open issues 🎉

1 previously missed issue

In code that hasn't changed since last review

Bound flattening accumulator to prevent oversized intermediate allocations

src/​jsonata/​functions.py:2055

This check runs after evaluate_wildcard has flattened each array into an ordinary list (src/jsonata/jsonata.py:709–710, 717–718). With sequence=2, a large input array is therefore copied in full before D2015 is raised, so the intermediate allocation can still exhaust memory. Make the accumulator in Jsonata.flatten bounded so it checks the limit as it grows.

1 resolved since last review

• Avoid enforcing sequence bounds after evaluation

_{Review effort: Balanced}