test: hold json_tree containment to jsonb with proptest, and restructure the JSON index tests by coderdan · Pull Request #1345 · cipherstash/stack

coderdan · 2026-10-23T16:05:00Z

Summary

EQL v4's json_tree column lets PostgreSQL answer JSON containment over encrypted documents: @> ("does this document contain that one?") and <@ (the reverse). The promise is that it answers exactly as PostgreSQL's own jsonb operators would answer over the plaintexts. Until now that promise was checked against live PostgreSQL only for a fixed, hand-picked set of documents. The only randomized check compared against a Rust model of jsonb containment, never against PostgreSQL itself.

This PR adds property tests built on proptest. Each test generates batches of documents and queries and checks every answer against PostgreSQL's real jsonb answer, on both the index plan and the sequential-scan plan. A failure shrinks to a small document and query. It is test-only.

It also restructures packages/stack-encrypt/tests/json_index.rs, the in-memory half of the same oracle (no database: it holds the encrypted projection to a Rust model of jsonb). Its tests now follow the repository's Rust testing conventions, and its hand-rolled random check is replaced by proptest properties like the ones above. No assertion was dropped; the old-to-new mapping is below.

Changes

packages/eql/tests/encryption/tests/json_tree/generated.rs (new, included from json_tree.rs as mod generated):
- What a case is: a batch of up to 64 stored documents and up to 96 needles (the queries). The default run is 64 cases.
- Document strategy: a recursive (prop_recursive) JSON value. It covers:
  - nested objects and arrays, with keys reused across levels and siblings
  - arrays of objects, arrays of arrays, and duplicate elements
  - numbers that are equal but written differently (1 / 1.0 / 1e0, 100 / 1e2, 0 / -0.0, 0.5 / 5e-1)
  - strings that look like numbers, booleans or null
  - booleans, null, and empty {} and []
  - root scalars and root arrays
- Needle strategies:
  - random values
  - prunings of a stored document: of the whole document, which that row contains by construction (checked against PostgreSQL), or of a node inside it
  - near misses: a pruning with one edit (a changed leaf, an added key, an added element, or a changed kind)
  A derived needle refers to its document by position and to its pruning by a keep mask. When proptest shrinks the documents, the needle still refers to a valid document, and the needle itself shrinks toward smaller.
- Oracle: each document is stored twice in the same row, encrypted in a public.eql_v4_json_tree column and as plaintext jsonb. A case is checked in a few set-based statements. Each statement answers every needle at once with one correlated subquery per needle. PostgreSQL computes the jsonb answer, and the encrypted answer must equal it for every needle.
- Plans: every case runs on both plans: sequential scans off (the GIN index), and every index path off. EXPLAIN confirms each plan. The table has no primary key, so the GIN index is the only index the planner can take.
- The three properties, as separate tests under given_generated_documents_in_postgres:
  - contains_selects_the_rows_jsonb_contains_on_both_plans (@>)
  - contained_by_selects_the_rows_jsonb_does_except_a_root_scalar_in_a_root_array (<@). The only allowed difference is the documented one (a stored root scalar inside a needle root array). The test asserts that this is the whole difference and that the run hits the case at least once.
  - arrow_opens_a_leaf_to_the_value_jsonb_holds (-> at 20 key paths). The encrypted column has an entry exactly where jsonb #> finds a node, and a leaf opens (decrypts) to the value jsonb holds there.
- Generator coverage: given_the_document_strategy::produces_every_feature_the_oracle_relies_on needs no database. It runs in test:encryption and fails if the generator stops producing any feature listed above.
- Structure: follows the Rust testing conventions: given_* scenario modules, test names that state the expectation, a scenario() helper, an assertions submodule, a message on every assertion, and #[allow(clippy::unwrap_used)] on the module.
packages/eql/tests/encryption/proptest-regressions/json_tree.txt: where a failing case's seed is recorded. Every later run replays it first. The file is set explicitly with FileFailurePersistence::Direct, because proptest's default location is resolved in a way that does not fit this nested crate.
packages/eql/tests/encryption/Cargo.toml, packages/eql/Cargo.lock: proptest = "1" as a dev-dependency. That is the same requirement eql-domains and tests/sqlx carry, so the lockfile only gains one dependency edge on the existing 1.11.0, not a second copy.
packages/eql/mise.toml: test:encryption:postgres passes --show-output, so the seed, case count and summary lines appear in the CI log even when the tests pass.
json_tree.rs: the hand-rolled generated test from this PR's first commit is replaced by the module above. The file's other tests are unchanged.
packages/stack-encrypt/tests/json_index.rs (rewritten, 28 tests → 47):
- Structure: one given_<scenario> module per situation, each test named for what it expects. A scenario() helper builds the cipher and plan. An assertions module holds the checks that repeat: shape-mismatch, tamper, term and two-context refusals, round trips, and both containment directions. Every assertion has a message. #![allow(clippy::unwrap_used)] sits at the crate root, which is the test module of an integration test.
- Splits: the longest tests are split along the expectations they made, and four near-identical assertions became a loop.
- What stayed hardcoded: the fixture document, the design corpus and the index's wire strings stay literal, because the tests pin entry ordinals and a wire contract. No test uses a database, so the skill's reason for randomised data (avoiding row-lock contention) does not apply.
- proptest: the Lcg random check becomes two properties, @> and <@, over one generated stored document and one needle per case. A needle is a random value, the document itself, a pruning of it, or a pruning with one edit. The oracle is the in-file jsonb model, which given_the_containment_model pins to PostgreSQL's answers.
  - Default run: a ChaCha RNG with seed 1336 and 512 cases per property. PROPTEST_RNG_SEED and PROPTEST_CASES override them.
  - Failures persist to packages/stack-encrypt/proptest-regressions/json_index.txt.
  - A census test fails if the document strategy stops producing a feature.
  - The strategies duplicate the EQL test's, because the two are in separate Cargo workspaces.
packages/stack-encrypt/Cargo.toml, Cargo.lock: proptest = { version = "1.7", default-features = false, features = ["std"] } as a dev-dependency, written the way stack-guest-abi writes it. It resolves to the 1.11.0 already in the root lockfile.
docs/plans/2026-10-17-json-tree-index.md: the description of both oracles now names the proptest properties. No file in the repo referred to the old test names or to json_index.rs:<line>.

Old → new test names in json_index.rs (28 → 47)

Old	New
`a_document_seals_one_entry_per_node_and_opens_again`	`given_a_document::seals_one_entry_per_node_and_opens_again`
`the_deepest_document_the_index_takes_opens_again`	`given_the_deepest_document::opens_again_and_one_level_deeper_is_refused_when_sealed`
`the_wire_form_is_what_the_sql_domain_stores`	`given_the_fixture_document::stores_the_wire_form_the_sql_domain_checks`
`the_chain_fields_sit_where_the_design_says`	`given_the_fixture_document::{places_the_chain_fields_where_the_design_says, gives_equal_values_at_one_path_one_value_selector_and_two_ciphertexts, chains_the_top_level_fields_from_an_open_root}`
`an_ordering_term_sits_only_at_an_ordered_path`	`given_ordered_paths::{puts_a_term_only_on_a_number_or_string_leaf_at_an_ordered_path, refuses_a_position_or_a_text_that_is_not_a_path}`
`a_token_derived_from_the_path_alone_unlocks_exactly_its_subtree`	`given_a_subtree_key::{unlocks_exactly_its_subtree_from_the_path_alone, gates_each_key_step_below_its_head, has_no_token_at_a_sealed_root_and_derives_from_an_open_one}`
`a_selector_is_the_stored_one_and_folds_positions_into_the_wildcard`	`given_the_fixture_document::derives_the_stored_selector_from_a_path_folding_positions_into_the_wildcard`
`an_extracted_node_opens_to_its_scalar_and_a_subtree_to_its_value`	`given_an_extracted_node::{opens_to_its_scalar_and_a_walked_subtree_to_its_value, round_trips_with_the_envelope_the_sql_grafts}`
`subtree_rows_the_server_repeats_or_lines_up_are_refused`	`given_rows_the_server_chose::refuses_a_repeated_row_or_a_line_of_containers`
`a_moved_reordered_or_regrafted_entry_fails_to_open`	`given_a_tampered_document::fails_to_open_a_moved_reordered_or_regrafted_entry`
`the_containment_model_gives_postgresqls_answers`	`given_the_containment_model::gives_postgresqls_pinned_answers`
`containment_over_the_projection_is_postgresqls_containment`	`given_the_design_corpus::answers_containment_both_ways_as_jsonb_does`; `given_generated_documents::{contains_as_jsonb_contains, is_contained_by_as_jsonb_except_a_root_scalar_in_a_root_array}`; `given_generated_documents::given_the_document_strategy::produces_every_feature_the_property_relies_on`
`a_query_term_serializes_as_the_needle_operand_and_derives_no_key`	`given_a_containment_query::serializes_as_the_needle_operand_and_derives_no_key`
`an_accessor_asks_at_a_path`	`given_an_accessor::{matches_the_value_at_its_path, orders_as_the_stored_leaf_at_an_ordered_path, refuses_a_bad_path_an_unordered_leaf_or_a_path_the_plan_does_not_order, orders_at_a_wildcard_and_refuses_a_position, refuses_a_container_for_equality}`
`a_laid_out_plan_opens_what_it_wrote`	`given_a_laid_out_plan::{reports_its_layout, opens_what_it_wrote_one_document_or_a_batch, refuses_an_open_that_names_another_context, answers_a_named_index_with_its_own_options}`
`equality_at_the_root_is_not_containment`	`given_equality_at_the_root::matches_the_root_value_and_not_an_array_holding_it`
`a_query_under_other_options_is_refused_by_the_plan`	`given_a_plan::refuses_a_query_under_other_options`
`opening_through_the_plan_is_the_typed_open`	`given_a_plan::opens_what_it_wrote_as_the_typed_open_does` (now also compares with the typed open)
`the_json_index_has_a_wire_form` (feature `dynamic`)	`given_the_json_index_spec::has_a_wire_form`
`no_stored_tag_is_a_token`	`given_the_fixture_document::stores_no_token`
`a_position_reaches_no_object_member`	`given_a_position::{reaches_no_object_member, gates_on_the_elements_selector_below_a_key_step}`
`gated_walk_rows_open_to_the_reached_nodes_value`	`given_gated_walk_rows::open_to_the_reached_nodes_value`
`each_stored_form_opens_as_a_record_field`	`given_a_derived_record::opens_each_stored_form_as_its_field`
`the_sealed_root_a_walk_fell_back_to_opens_at_the_path`	`given_a_fallback_root::{opens_at_the_keys_path, refuses_one_value_where_the_path_reaches_none_or_two, is_only_the_sealed_root_alone_at_depth_zero}`
`the_key_built_fallback_opens_to_what_the_walk_opens_to`	`given_a_fallback_root::opens_to_what_the_walk_opens_to`
`a_walked_row_carries_its_depth`	`given_an_extracted_node::carries_no_depth_until_a_walk_gives_it_one`
`a_document_without_the_sealed_root_opens_from_its_entries`	`given_an_open_root::keeps_the_root_as_an_entry_and_opens_from_the_entries`
`an_accessor_is_refused_before_its_keyset_loads`	`given_an_accessor::is_refused_before_its_keyset_loads`

Encryption uses the test crate's fake key source, so no CipherStash credentials are needed.

Replaying a failure

The run is deterministic: a ChaCha RNG with seed 1336, 64 cases. Every run logs, for example:

json_tree generated @>: seed 1336, 64 cases; replay with PROPTEST_RNG_SEED=1336 PROPTEST_CASES=64

PROPTEST_RNG_SEED and PROPTEST_CASES override the seed and the case count, the standard proptest variables. A failure prints the shrunk case: each stored document and needle as JSON, the two answers, and the rows they disagree on. It also appends the case's seed to proptest-regressions/json_tree.txt. Commit that line with the fix.

json_index.rs works the same way, with seed 1336 and 512 cases. For example:

PROPTEST_RNG_SEED=42 mise x --env test -- cargo nextest run -p stack-encrypt --all-features --test json_index -- given_generated_documents

A failure prints the minimal stored document and needle, and appends the seed to packages/stack-encrypt/proptest-regressions/json_index.txt.

Verification

Locally, against the EQL tasks' own Postgres (mise run postgres:up postgres-N, port 7432), with the v4 bundle built from this tree:

mise run test:encryption:postgres passes on PostgreSQL 17 and 14.24. The json_tree suite took 15.4 s on 17 and 16.5 s on 14. The three properties run in parallel; the slowest is <@, at about 15 s.
Default seed counts: 3,345 needles asked, 1,329 of 1,352 near misses missing their row, 172 rows excluded under the <@ exception, and 254 leaves opened and 1,957 containers found by ->.
Other seeds, all passing:
- PG 17: 1, 2, 42, 777, 9001, 31337, 123456789, 18446744073709551615
- PG 14: 3, 2024, 65535
- PG 17, 512 cases with seed 20261024: 24,463 needles, about 107 s
I broke the <@ exception on purpose. The test failed and shrank to one stored document 0 and the needle [0], which is exactly the documented case. The change was reverted, along with the regression line it wrote.
cargo fmt --check and the task's clippy lines (-p eql-bindings -p eql-encryption-tests --all-features --all-targets -D warnings, and -p eql-bindings --features stack-encrypt) are clean. cargo test --locked -p eql-encryption-tests passes.
All repo script tests pass (vitest --config scripts/vitest.config.mjs: 1,371 passed, 29 skipped), including the ones that parse packages/eql/mise.toml, the Cargo manifests and the workflows. All 78 files are vitest suites, so node --test does not apply to them.

For stack-encrypt:

mise x --env test -- cargo nextest run --locked -p stack-encrypt --all-features: 784 passed, 1 skipped, including all 47 json_index tests and the trybuild ui binary. The two properties take about 0.76 s for 512 cases.
cargo test -p stack-encrypt --no-default-features --test json_index (the shape wasm:no-http-test builds) passes 46 tests. The dynamic-only wire-form test is left out there.
cargo fmt --all --check and the workflow's cargo clippy --locked --no-deps --workspace --all-targets --all-features -- -D warnings are clean.
Seeds 1, 2, 42, 9001 and 18446744073709551615 pass, as does 4,096 cases with seed 20261024 (6 s).
I broke the <@ exception in the property on purpose. It shrank to stored 0 and needle [0], and was then reverted.

CI: run 40047240427 is green, ci-required included. In Rust encryption with PostgreSQL 17, all three properties passed (-> in 4.6 s, @> in 11.4 s, <@ in 15.1 s; the suite took 15.1 s). The log shows the same seed and counts as the local run. The database-free census test passed in Rust workspace crates. On a pull request, test-eql.yml runs PostgreSQL 17 only. Versions 14–16 run on a push to main and nightly, so for this PR PG 14 was checked locally only (above). The first commit's version passed CI on PG 17 in run 39983784023, and locally on 14, 15 and 16.

CI for the json_index commit (f2e998c): all 48 checks pass. The 2 skipped are the nightly fuzz campaign and the Claude review, which this event does not run. In Tests (crates) (fmt, clippy, nextest, doctests, rustdoc), nextest ran all 47 json_index tests. The Mutants gate reported no mutants to test, since the diff touches no src, and passed. Unused dependencies, CRAP, Fuzz regression and Miri also pass, and so does Test EQL (ci-required).

No changeset: test-only.

Review notes

The skill's runner convention (nextest) is not applied. test:encryption:postgres runs cargo test -- --ignored behind a listing guard that other suites share, and changing the runner is outside this PR.
json_index.rs duplicates the EQL generator rather than sharing it, because the two are separate Cargo workspaces. Its properties hold the projection to the in-file Rust model of jsonb; the EQL tests are the ones that check against PostgreSQL itself.
Only json_index.rs was restructured. The other stack-encrypt test files and tests/common are unchanged.

The live json_tree oracle checked a fixed corpus and a fixed list of needles. The only randomized containment check compared the projection with an in-memory model of jsonb containment, never with PostgreSQL, so nothing generated was ever answered by the database itself. postgres_containment_over_generated_documents_is_jsonb_containment stores 300 generated documents encrypted beside their plaintexts and asks 500 needles: random values, sub-documents pruned from stored documents, and those sub-documents with one edit. For every needle the rows the encrypted @> and <@ select must be the rows jsonb @> and <@ select over the plaintexts, as PostgreSQL computes them. It runs once with the planner held to the GIN index and once with every index path off, and checks each @> plan with EXPLAIN. <@ may differ only where the SQL says it does (a stored root scalar inside a needle root array), and the test asserts that is the whole difference. It also checks -> at 20 generated key paths: the entry exists exactly where jsonb #> finds a node, and a leaf opens to the value jsonb holds there. The seed is fixed (1336), logged on every run, and replaced by EQL_JSON_TREE_SEED. A failure prints the seed, the needle and every stored document the two answers disagree on. test:encryption:postgres passes --show-output so the seed line reaches the CI log. Claude-Session: https://claude.ai/code/session_01XQN1QpaWVDGjHEg39KEJF3

changeset-bot · 2026-10-23T16:05:06Z

⚠️ No Changeset found

Latest commit: f35ee53

Merging this PR will not cause a version bump for any packages. If these changes should not result in a new version, you're good to go. If these changes should result in a version bump, you need to add a changeset.

This PR includes no changesets

When changesets are added to this PR, you'll see the packages that this PR includes changesets for and the associated semver types

Click here to learn what changesets are, and how to add one.

Click here if you're a maintainer who wants to add a changeset to this PR

coderabbitai · 2026-10-23T16:05:10Z

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

Run configuration

Source mode: Auto
Repository: Unavailable
Branch: Unavailable
File: No configuration file from this repository applied
Commit: Unavailable
Layers: Organization UI; Enforced overrides: Organization
Review profile: CHILL
Plan: Advanced
Run ID: 5d64098a-0882-42eb-ac6f-9d3af0b7bc57

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

🔍 Trigger review

Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

_{Comment @coderabbitai help to get the list of available commands.}

The generated containment oracle drew its documents and needles from a hand-rolled SplitMix64 generator, one fixed batch per run, and could not shrink a failure. It now draws them from proptest strategies: a prop_recursive JSON value with the same coverage (nested objects and arrays, repeated keys, arrays of objects and of arrays, duplicate elements, numbers equal by value in different representations, numeric strings, booleans, null, empty containers, root scalars and arrays), and needles that are random values, prunings of a stored document, or such a pruning with one edit. A derived needle names its document by an Index and its pruning by a keep mask, so it follows the documents as they shrink and shrinks toward a smaller needle itself. A fault injected into the <@ exception shrinks to one stored 0 and the needle [0]. A case is a batch of documents and needles, checked in a few set-based statements: one correlated subquery per needle, which the planner serves from the GIN index when sequential scans are off. The table has no primary key, so the GIN index is the only one there is to take. Every case asserts with EXPLAIN that each plan is the one it asked for. The three properties (@> on both plans, <@ on both plans less its documented exception, and -> at generated key paths) are separate tests under given_generated_documents_in_postgres, and a database-free test holds the document strategy to every feature the census counts. The module lives in tests/json_tree/generated.rs. The RNG is ChaCha with seed 1336 and 64 cases by default. PROPTEST_RNG_SEED and PROPTEST_CASES override them, as for any proptest run, and both are logged. Failures persist to the crate's proptest-regressions/json_tree.txt through FileFailurePersistence::Direct, because SourceParallel resolves file!() from the working directory, which is not where this nested workspace member's paths are rooted. Claude-Session: https://claude.ai/code/session_01XQN1QpaWVDGjHEg39KEJF3

coderdan · 2026-10-24T01:18:47Z

Reworked in dd9bc51 (pushed as a new commit, no history rewrite):

proptest replaces the hand-rolled generator. A recursive strategy generates the JSON documents with the same coverage as before. Needles are random values, prunings of a stored document, or near misses (a pruning with one edit). They refer to their stored document by position, so a failure shrinks to a small document and needle. A fault I injected into the <@ exception shrank to one stored 0 and the needle [0].
One case = a batch (up to 64 documents, up to 96 needles), checked in a few set-based statements against live PostgreSQL. Every case runs on both plans, the GIN index and a sequential scan, and EXPLAIN confirms each one. Default run: 64 cases.
Deterministic: a ChaCha RNG with seed 1336. PROPTEST_RNG_SEED and PROPTEST_CASES override it, and the log prints the exact replay line. Failure seeds go to packages/eql/tests/encryption/proptest-regressions/json_tree.txt.
Restructured to the Rust testing conventions: given_* modules and three expectation-named properties (@>, <@, ->), plus a database-free test that fails if the generator stops producing any feature.
Runtime: 15.1 s on PG 17 in CI and 16.5 s on PG 14 locally. No mismatch on any seed tried, including a 512-case run.

json_index.rs's in-memory test is left alone; the reason is under Review notes in the description.

https://claude.ai/code/session_01XQN1QpaWVDGjHEg39KEJF3

chatgpt-codex-connector · 2026-10-24T01:20:15Z

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-24T01:21:28.034869Z	`dd9bc51`	Draft marked ready
🔒 Security Review	✅ Completed 2026-10-24T01:22:09.185799Z	`dd9bc51`	Draft marked ready

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

…ntainment with proptest tests/json_index.rs held 28 flat tests named for whole sentences, with bare assertions and a hand-rolled Lcg for its only randomized check. It now follows the repository's Rust testing conventions: one given_<scenario> module per situation and a test named for what it expects, a scenario() helper for the cipher and plan, an assertions module for the refusals and round trips the tests repeat, a message on every assertion, and unwrap allowed at the crate root, which is the test module of an integration test. No assertion was dropped. The longest tests were split along the expectations they made (an accessor's matching, ordering and refusals are five tests now; a subtree key's derivation, gates and root handling are three), and four near-identical assertions became loops. The run goes from 28 tests to 47. The random containment check becomes two proptest properties, one for @> and one for <@, over a generated stored document and a needle: a random value, the document itself, a pruning of it, or a pruning with one edit. They hold the projection to the in-file model of jsonb, which given_the_containment_model pins to PostgreSQL's answers. A failure shrinks to a small pair: breaking the <@ exception shrinks to stored 0 and the needle [0]. The RNG is ChaCha, seed 1336, 512 cases per property. PROPTEST_RNG_SEED and PROPTEST_CASES replace them. Failures persist to the crate's proptest-regressions/json_index.txt. A census test fails if the document strategy stops producing a feature the property relies on. The strategy duplicates the EQL live oracle's, which sits in another Cargo workspace. proptest is a dev-dependency with default features off, as stack-guest-abi takes it, resolving to the 1.11.0 already in the lock. The plan document's description of both oracles is updated to match. Claude-Session: https://claude.ai/code/session_01XQN1QpaWVDGjHEg39KEJF3

github-actions · 2026-10-24T01:39:27Z

Mutation testing (cargo-mutants, `--in-diff`, stack-auth + stack-encrypt)

No mutants were generated for the changed lines.

coderdan · 2026-10-24T01:54:08Z

Added f2e998c, which restructures packages/stack-encrypt/tests/json_index.rs (new commit, no history rewrite):

Structure: the file now follows the Rust testing conventions. It has given_<scenario> modules with tests named for what they expect, a scenario() helper, an assertions module, a message on every assertion, and #![allow(clippy::unwrap_used)]. The 28 tests are now 47 because long tests were split along their expectations. No assertion was dropped; the old-to-new mapping is in the description.
proptest: the Lcg random check is now two proptest properties (@> and <@), with a fixed ChaCha seed of 1336 and 512 cases each. PROPTEST_RNG_SEED and PROPTEST_CASES override them, and failures persist to packages/stack-encrypt/proptest-regressions/json_index.txt. A census test guards the generator. A fault I injected shrank to stored 0 and needle [0].
What stayed hardcoded: the fixture document, the design corpus and the wire strings, because the tests pin entry ordinals and a wire contract.
CI: all green. Tests (crates) ran all 47 under nextest, and the mutants gate found nothing to mutate (the diff is test-only).

https://claude.ai/code/session_01XQN1QpaWVDGjHEg39KEJF3

…roperties Brings in the rename of the plan-layer Plan::value().encrypt_as(layout) to encrypt_with_layout(layout), the generic-plan assertions in the planner test and the row-provenance wording fix. The one conflict was in packages/stack-encrypt/tests/json_index.rs, which this branch rewrites: the rewritten file is kept and the rename applied to its two plan-layer call sites. The cipher-layer encrypt_as calls are unchanged. Claude-Session: https://claude.ai/code/session_01N7bCqa3ZVJMcEE99RzVqEu

…roperties

…roperties Cargo.lock conflict only: the base branch's lock, which now carries next, plus this branch's proptest edge. Claude-Session: https://claude.ai/code/session_01N7bCqa3ZVJMcEE99RzVqEu

…roperties

This was referenced Oct 24, 2026

Tracking: encrypted JSON documents (json_tree) in the Go SDK #1347

Open

docs(plans): the json_tree capability in the Go SDK #1353

Draft

coderdan marked this pull request as ready for review October 24, 2026 01:20

coderdan requested a review from a team as a code owner October 24, 2026 01:20

coderdan changed the title ~~test(eql): hold json_tree containment to jsonb over generated documents~~ test: hold json_tree containment to jsonb with proptest, and restructure the JSON index tests Oct 24, 2026

coderdan added 6 commits October 24, 2026 14:40

Merge branch 'feat/json-tree-gates' into test/json-tree-containment-p…

7d1f9e5

…roperties

Merge branch 'feat/json-tree-gates' into test/json-tree-containment-p…

0f80f8c

…roperties

Merge branch 'feat/json-tree-gates' into test/json-tree-containment-p…

f2af67c

…roperties Cargo.lock conflict only: the base branch's lock, which now carries next, plus this branch's proptest edge. Claude-Session: https://claude.ai/code/session_01N7bCqa3ZVJMcEE99RzVqEu

Merge branch 'feat/json-tree-gates' into test/json-tree-containment-p…

338e77b

…roperties

Merge branch 'feat/json-tree-gates' into test/json-tree-containment-p…

f35ee53

…roperties

Base automatically changed from feat/json-tree-gates to next October 24, 2026 06:12

Navigation Menu

Repository navigation

test: hold json_tree containment to jsonb with proptest, and restructure the JSON index tests - #1345

test: hold json_tree containment to jsonb with proptest, and restructure the JSON index tests#1345
coderdan wants to merge 9 commits into
nextfrom
test/json-tree-containment-properties

coderdan commented Oct 23, 2026 •

edited

Loading

Uh oh!

changeset-bot Bot commented Oct 23, 2026 •

edited

Loading

Uh oh!

coderabbitai Bot commented Oct 23, 2026 •

edited

Loading

Review skipped

Uh oh!

coderdan commented Oct 24, 2026

Uh oh!

chatgpt-codex-connector Bot commented Oct 24, 2026 •

edited

Loading

Uh oh!

github-actions Bot commented Oct 24, 2026

Uh oh!

coderdan commented Oct 24, 2026

Uh oh!

This branch has not been deployed

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

coderdan commented Oct 23, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Changes

Replaying a failure

Verification

Review notes

Related

Uh oh!

changeset-bot Bot commented Oct 23, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

⚠️ No Changeset found

Uh oh!

coderabbitai Bot commented Oct 23, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Review skipped

Uh oh!

coderdan commented Oct 24, 2026

Uh oh!

chatgpt-codex-connector Bot commented Oct 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codex Review Summary

Uh oh!

github-actions Bot commented Oct 24, 2026

Mutation testing (cargo-mutants, --in-diff, stack-auth + stack-encrypt)

Uh oh!

coderdan commented Oct 24, 2026

Uh oh!

This branch has not been deployed

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

coderdan commented Oct 23, 2026 •

edited

Loading

changeset-bot Bot commented Oct 23, 2026 •

edited

Loading

coderabbitai Bot commented Oct 23, 2026 •

edited

Loading

chatgpt-codex-connector Bot commented Oct 24, 2026 •

edited

Loading

Mutation testing (cargo-mutants, `--in-diff`, stack-auth + stack-encrypt)