Datasets:
instance_id stringlengths 12 60 | repo stringclasses 927
values | pr_number int64 3 61.3k | language stringclasses 19
values | base_commit stringlengths 40 40 | variant stringclasses 3
values | problem_statement stringlengths 59 5.71k | generation stringclasses 2
values |
|---|---|---|---|---|---|---|---|
0xpolygonhermez__zkevm-node-1321 | 0xPolygonHermez/zkevm-node | 1,321 | go | 63a0490d5e7e87c230aece5c6281da3063ba6654 | narrative | ## Task: Optional Read-Only Mode for Etherman
### Intent
Introduce an optional read-only operating mode for the Etherman component so that nodes which never submit transactions (such as read-only followers or RPC-serving nodes in zkEVM/Polygon Hermez deployments) can be configured and started without provisioning an E... | model |
0xpolygonhermez__zkevm-node-1321 | 0xPolygonHermez/zkevm-node | 1,321 | go | 63a0490d5e7e87c230aece5c6281da3063ba6654 | acceptance | ## Task: Read-Only Mode for Etherman
### Done when...
1. The Etherman config struct exposes a new boolean field `ReadOnly` that defaults to `false`, and existing node configs that omit it continue to behave exactly as before (backward compatible).
2. When `ReadOnly` is `false` (the default), the Etherman instance is c... | model |
0xpolygonhermez__zkevm-node-1321 | 0xPolygonHermez/zkevm-node | 1,321 | go | 63a0490d5e7e87c230aece5c6281da3063ba6654 | scope | ## Task: Read-Only Mode for Etherman
### In scope
- **Etherman package:**
- Modify struct construction so that when `ReadOnly` is `true`, the instance is built without an Ethereum account and the account-loading step is skipped.
- Add an internal read-only state flag (e.g., a `readOnly` field on the struct) and an... | model |
0xs34n__starknet.js-490 | 0xs34n/starknet.js | 490 | ts | e0d22397eac39ec60ca58f2f7e420fea63047ecf | narrative | Relocate the `simulateTransaction` capability onto the `Account` class so that simulation is always performed with an account signature, matching the existing convention used by `estimateFee`. Once the move is complete, any invocation of the method through the Provider/Contract layer must fail with an explicit error ra... | model |
0xs34n__starknet.js-490 | 0xs34n/starknet.js | 490 | ts | e0d22397eac39ec60ca58f2f7e420fea63047ecf | acceptance | Acceptance criteria — done when ALL of the following hold:
1. Given an `Account` instance, calling `account.simulateTransaction(invocation, ...)` with the required signature returns the simulated transaction result without throwing.
2. Given the previous Provider/Contract call path, invoking `simulateTransaction` (or ... | model |
0xs34n__starknet.js-490 | 0xs34n/starknet.js | 490 | ts | e0d22397eac39ec60ca58f2f7e420fea63047ecf | scope | Scope and boundaries:
In scope:
- Move the `simulateTransaction` method (along with its request/response types and response parser) from its current location to the `Account` class.
- Make `Account.simulateTransaction` require the account-signature context that `Account` provides, matching the existing `estimateFee` p... | model |
aeon-toolkit__aeon-2822 | aeon-toolkit/aeon | 2,822 | python | 55188f94c6f88e12e851de13581a4d30bfc9a07a | narrative | ## Intent
Migrate the continuous anomaly-detection metric helpers in `aeon/anomaly_detection/metrics/_continuous.py` off the external `prts` package and onto the in-repo range-based metrics implementation introduced for issue #2785. The deprecated `p_alpha` parameter should disappear from these functions, and the `alp... | model |
aeon-toolkit__aeon-2822 | aeon-toolkit/aeon | 2,822 | python | 55188f94c6f88e12e851de13581a4d30bfc9a07a | acceptance | ## Done When…
1. `aeon/anomaly_detection/metrics/_continuous.py` contains no import (direct or transitive for metric computation) from the `prts` package; every continuous metric function delegates to the in-repo range-based metrics implementation already present in the repository.
2. The public function signatures of... | model |
aeon-toolkit__aeon-2822 | aeon-toolkit/aeon | 2,822 | python | 55188f94c6f88e12e851de13581a4d30bfc9a07a | scope | ## In Scope
- `aeon/anomaly_detection/metrics/_continuous.py` — the continuous anomaly-detection metric function implementations. Each function must stop importing from the external `prts` package and must instead delegate to the in-repo range-based metrics implementation (the one referenced by issue #2785). The depre... | model |
aeye-lab__pymovements-1106 | aeye-lab/pymovements | 1,106 | python | f32eaa09e889a450aa9123bbf8ad5e324ca87753 | narrative | In the EyeLink ASC parser, the screen resolution is currently inferred from the optional `DISPLAY_COORDS` line. Because that line is not present in every recording, files without it end up with no resolution metadata and downstream processing fails. Switch the resolution source to `GAZE_COORDS`, which is emitted reliab... | model |
aeye-lab__pymovements-1106 | aeye-lab/pymovements | 1,106 | python | f32eaa09e889a450aa9123bbf8ad5e324ca87753 | acceptance | Done when:
1. Given an ASC file containing one or more `GAZE_COORDS` messages, the parser extracts the screen resolution from those messages and exposes it inside `recording_config` metadata.
2. Given an ASC file that has `GAZE_COORDS` but no `DISPLAY_COORDS`, the parser still produces a non-null screen resolution.
3. ... | model |
aeye-lab__pymovements-1106 | aeye-lab/pymovements | 1,106 | python | f32eaa09e889a450aa9123bbf8ad5e324ca87753 | scope | Explicit boundaries, then the work.
In scope:
- In `parse_eyelink()`, change the source from which screen resolution is read from the optional `DISPLAY_COORDS` message to the `GAZE_COORDS` message.
- Place the parsed resolution into the `recording_config` metadata for the relevant recording block.
- Compare the resolu... | model |
aeye-lab__pymovements-408 | aeye-lab/pymovements | 408 | python | fd650c35b00e986ef52c3b2058fe7c3b7804d3d1 | narrative | ### Intent
Introduce an optional `name` argument to every event detection callable exposed by the dataset interface. When supplied, this argument overrides the algorithm-derived default identifier used for the detected events in the resulting dataset output.
### Motivation
When practitioners run several event detectio... | model |
aeye-lab__pymovements-408 | aeye-lab/pymovements | 408 | python | fd650c35b00e986ef52c3b2058fe7c3b7804d3d1 | acceptance | ### Done When...
1. Calling `dataset.detect_events('idt', name='fixation.idt')` produces events whose identifier in the dataset output is `fixation.idt`.
2. Calling `dataset.detect_events('ivt', name='fixation.ivt')` produces events whose identifier in the dataset output is `fixation.ivt`.
3. Calling `dataset.detect_ev... | model |
aeye-lab__pymovements-408 | aeye-lab/pymovements | 408 | python | fd650c35b00e986ef52c3b2058fe7c3b7804d3d1 | scope | ### Scope
This task covers:
- Adding an optional `name` keyword parameter to every event detection function/method exposed through the dataset.
- Wiring the supplied value through to the events dataframe so the resulting event columns/identifiers in the dataset output reflect the user-supplied label.
- Adding one unit ... | model |
aeye-lab__pymovements-767 | aeye-lab/pymovements | 767 | python | 84f1416111b105f96f9270edb15c72311409136f | narrative | Extend the pymovements ASC parsing pipeline so that researchers can recover session-level metadata embedded in their EyeLink recordings. Currently, `from_asc()` and the lower-level `parse_eyelink()` helper only return gaze samples, silently discarding lines such as recording date, display resolution, and other custom f... | model |
aeye-lab__pymovements-767 | aeye-lab/pymovements | 767 | python | 84f1416111b105f96f9270edb15c72311409136f | acceptance | Done when:
1. `from_asc()` accepts a new optional keyword argument `metadata_patterns`, whose value is a dict mapping user-defined keys to regex pattern strings.
2. `parse_eyelink()` accepts the same `metadata_patterns` argument with identical semantics.
3. After parsing, both functions return a tuple `(gaze_data, met... | model |
aeye-lab__pymovements-767 | aeye-lab/pymovements | 767 | python | 84f1416111b105f96f9270edb15c72311409136f | scope | Scope of this task:
IN SCOPE:
- Adding an optional `metadata_patterns` keyword argument to `from_asc()` and to `parse_eyelink()` in the pymovements library.
- Extracting session-level metadata from EyeLink ASC files using the supplied regex patterns and returning it as a dictionary.
- Changing the return signature of ... | model |
agnostiqhq__covalent-1377 | AgnostiqHQ/covalent | 1,377 | python | d5cb36a03fb8f3f49c75d958d5a437901b7ea7df | narrative | ## Intent
Extend the Covalent S3 file transfer plugin with first-class folder (directory) upload and download support so that users can move entire directory trees to and from S3 buckets, not only single objects.
## Motivation
Today the S3 file transfer plugin handles individual files correctly, but directory transfer... | model |
agnostiqhq__covalent-1377 | AgnostiqHQ/covalent | 1,377 | python | d5cb36a03fb8f3f49c75d958d5a437901b7ea7df | acceptance | ## Done when (acceptance criteria)
1. Given a local folder, calling the upload-folder callable with an S3 destination bucket uploads the folder's contents to that bucket and does not raise a `404 HeadObject` error.
2. Given a remote folder in an S3 bucket, calling the corresponding download-folder callable retrieves th... | model |
agnostiqhq__covalent-1377 | AgnostiqHQ/covalent | 1,377 | python | d5cb36a03fb8f3f49c75d958d5a437901b7ea7df | scope | ## In scope
This task covers exactly one thing: adding folder (directory) upload and folder (directory) download support to the Covalent S3 file transfer plugin.
Specifically in scope:
- New folder upload callable in the S3 file transfer plugin that, given a local folder and an S3 destination bucket, transfers the fol... | model |
agnostiqhq__covalent-1460 | AgnostiqHQ/covalent | 1,460 | python | 17c0b8b7f156b3f04875c547f026ec4d87f4388e | narrative | Update the client-side `get_result` function so that it can retrieve computed outputs from a remote Covalent dispatcher without crashing, and pin the `networkx` dependency to a known-good version.
Motivation: Right now, when a user calls `get_result` against a dispatch id that lives on a remote Covalent server, the ca... | model |
agnostiqhq__covalent-1460 | AgnostiqHQ/covalent | 1,460 | python | 17c0b8b7f156b3f04875c547f026ec4d87f4388e | acceptance | Done when:
1. The client-side `get_result` function's signature includes a dispatcher-address argument (e.g. `get_result(<dispatch_id>, <dispatcher_address>)` or an equivalent updated signature).
2. Given a valid dispatch id and a remote dispatcher address, `get_result` returns the result without raising AttributeError... | model |
agnostiqhq__covalent-1460 | AgnostiqHQ/covalent | 1,460 | python | 17c0b8b7f156b3f04875c547f026ec4d87f4388e | scope | This task is a tightly scoped client-side bugfix. Being explicit about boundaries up front because we've had confusion before.
In scope:
- Modifying the client-side `get_result` function to accept a dispatcher-address argument so it can retrieve results from a remote Covalent server without raising AttributeError.
- P... | model |
agnostiqhq__covalent-1493 | AgnostiqHQ/covalent | 1,493 | python | ffad85610ec397918448d488021ebfd9ee09fe1d | narrative | ## Intent
Introduce a re-dispatch capability in the Covalent workflow orchestrator by exposing a new public helper, `ct.redispatch`, that lets users re-submit a workflow that has already been dispatched. The helper must accept options controlling three behaviors: retrying operations that previously failed, swapping out... | model |
agnostiqhq__covalent-1493 | AgnostiqHQ/covalent | 1,493 | python | ffad85610ec397918448d488021ebfd9ee09fe1d | acceptance | ## Acceptance criteria (done when...)
1. `ct.redispatch(handle)` re-submits the referenced workflow through the existing dispatcher and returns a result consistent with the original dispatch flow.
2. `ct.redispatch` accepts the existing dispatch ID/handle as its primary input.
3. A re-dispatch call can request retry of... | model |
agnostiqhq__covalent-1493 | AgnostiqHQ/covalent | 1,493 | python | ffad85610ec397918448d488021ebfd9ee09fe1d | scope | ## In scope
- A new public helper, `ct.redispatch`, that resubmits a previously dispatched workflow through the existing dispatcher using the original dispatch handle/ID.
- Three independent re-dispatch options:
- retry of operations that previously failed;
- replacement of specified electrons with new implementati... | model |
agnostiqhq__covalent-1508 | AgnostiqHQ/covalent | 1,508 | python | 30251684ac1d4d9783ca88f1cc2fc39f8f7f2275 | narrative | Covalent needs a way to re-submit an electron that was executed on another host without requiring the original function definition to be present locally. Currently, a user who wishes to rerun a previously dispatched electron must redefine the task by hand, which prevents portable and recoverable execution across machin... | model |
agnostiqhq__covalent-1508 | AgnostiqHQ/covalent | 1,508 | python | 30251684ac1d4d9783ca88f1cc2fc39f8f7f2275 | acceptance | Acceptance criteria (done when...):
1. Given a dispatch_id of a previously dispatched electron, calling `ct.redispatch()` on a host that can fetch that electron's result reconstructs an electron object equivalent to the original — same callable and same bound arguments.
2. The redispatched electron is submitted through... | model |
agnostiqhq__covalent-1508 | AgnostiqHQ/covalent | 1,508 | python | 30251684ac1d4d9783ca88f1cc2fc39f8f7f2275 | scope | In scope for this task:
1. Adding a public `ct.redispatch()` entry point that, given a reference to a previously dispatched electron on a host able to fetch the persisted result, reconstructs the original electron (function + arguments) from that result and re-submits it through the existing dispatcher pipeline, produc... | model |
agnostiqhq__covalent-1554 | AgnostiqHQ/covalent | 1,554 | python | 8c53ed187081ecba852572923d8b52a08f99740a | narrative | Intent: Introduce a postprocessing electron type in Covalent so that workflow post-processing is submitted to the runner through the same dispatch mechanism used for ordinary electrons, replacing the current dedicated workflow_executor post-processing path.
Motivation: Today the runner must handle two separate kinds o... | model |
agnostiqhq__covalent-1554 | AgnostiqHQ/covalent | 1,554 | python | 8c53ed187081ecba852572923d8b52a08f99740a | acceptance | Done when:
1. A workflow whose final step is post-processing executes that step as a postprocessing electron submitted to the runner rather than via a dedicated workflow_executor post-processing path.
2. The postprocessing electron exposes the same interface contract (inputs, outputs, executor binding, plugin hooks) as... | model |
agnostiqhq__covalent-1554 | AgnostiqHQ/covalent | 1,554 | python | 8c53ed187081ecba852572923d8b52a08f99740a | scope | Scope: introduce a postprocessing electron type in Covalent so that workflow post-processing is dispatched to the runner as a regular electron, eliminating the separate workflow_executor post-processing path. This is the change called for by linked issue #1545 and is justified by the principle that "all electrons look ... | model |
agnostiqhq__covalent-1617 | AgnostiqHQ/covalent | 1,617 | python | 0e67dfaf48954ed212e1052da24726a69028ce54 | narrative | Intent: Introduce a new `covalent deploy` CLI command group with three subcommands — `up`, `down`, and `status` — that allow users to provision, tear down, and inspect the cloud resources backing a named Covalent executor. Alongside the CLI, add a base cloud resource manager class that concrete per-provider managers (e... | model |
agnostiqhq__covalent-1617 | AgnostiqHQ/covalent | 1,617 | python | 0e67dfaf48954ed212e1052da24726a69028ce54 | acceptance | Done when:
1. Running `covalent deploy status <executor>` prints a human-readable summary of the executor's current resource state (for example: ready, provisioning, or failed) — not a raw object dump.
2. Running `covalent deploy up <executor>` is accepted by the CLI and routed to the corresponding cloud resource mana... | model |
agnostiqhq__covalent-1617 | AgnostiqHQ/covalent | 1,617 | python | 0e67dfaf48954ed212e1052da24726a69028ce54 | scope | In scope:
- Adding a new CLI command group `covalent deploy` with subcommands `up`, `status`, and `down`.
- Implementing the CLI wiring so that:
- `covalent deploy up <executor>` is accepted and routed to the corresponding cloud resource manager to initiate provisioning.
- `covalent deploy status <executor>` prints... | model |
agnostiqhq__covalent-1620 | AgnostiqHQ/covalent | 1,620 | python | 00afedf7aec33549cb334695a43c62b06eb5f27e | narrative | Intent: Add a centralized CloudResourceManager (CRM) to Covalent that owns cloud resource orchestration. The manager should provide a unified interface for provisioning and decommissioning compute resources, checking that required environment tools are present on the host, validating caller-supplied options, reading an... | model |
agnostiqhq__covalent-1620 | AgnostiqHQ/covalent | 1,620 | python | 00afedf7aec33549cb334695a43c62b06eb5f27e | acceptance | Done when — the following acceptance criteria are all true:
1. A CloudResourceManager class exists in the Covalent codebase and can be instantiated.
2. Given invalid user-supplied options, the manager raises a validation error rather than producing undefined behavior.
3. Given valid user-supplied options, the manager ... | model |
agnostiqhq__covalent-1620 | AgnostiqHQ/covalent | 1,620 | python | 00afedf7aec33549cb334695a43c62b06eb5f27e | scope | Scope-first problem statement for the CloudResourceManager (CRM) landing in Covalent.
This task covers:
- A new CloudResourceManager class that centralizes cloud resource orchestration.
- Public operations on the manager for: provisioning cloud resources for a target backend, tearing down previously provisioned cloud ... | model |
agnostiqhq__covalent-1660 | AgnostiqHQ/covalent | 1,660 | python | f513222ecf8d02500211fee994a04dadfd4a52ba | narrative | ## Intent
Make the `executor` attribute on an Electron instance mutable through direct property assignment, so callers no longer need to invoke a dedicated setter method to swap an Electron's executor after the object has been constructed, predefined, or imported.
## Motivation
This addresses issue #1647. The curren... | model |
agnostiqhq__covalent-1660 | AgnostiqHQ/covalent | 1,660 | python | f513222ecf8d02500211fee994a04dadfd4a52ba | acceptance | ## Done when…
1. The `Electron` class exposes a readable and writable `executor` property.
2. Assigning a value to `electron.executor` persists that value on the Electron instance.
3. Reading `electron.executor` returns the currently assigned executor.
4. Given a pre-defined or imported Electron, the new property acce... | model |
agnostiqhq__covalent-1660 | AgnostiqHQ/covalent | 1,660 | python | f513222ecf8d02500211fee994a04dadfd4a52ba | scope | ## In scope
- Modifying the `Electron` class to expose a read/write `executor` property.
- Updating the existing unit tests so they cover the new getter/setter behavior (both read and write paths).
- Updating the project documentation that describes Electron configuration so it reflects the new property usage.
- Addin... | model |
aio-libs__aiohttp-10300 | aio-libs/aiohttp | 10,300 | python | 3d3fba23cf543f7185658f52097f21093bf64c7c | narrative | Intent: Extend the top-level `aiohttp.client.request` function so it can receive keyword arguments drawn from `_RequestOptions` (for example `ssl`) and pass them down to the internal request handling path used by the rest of the client.
Motivation: Today the signature of `aiohttp.client.request` only enumerates a smal... | model |
aio-libs__aiohttp-10300 | aio-libs/aiohttp | 10,300 | python | 3d3fba23cf543f7185658f52097f21093bf64c7c | acceptance | Done when...
1. Calling `aiohttp.client.request(method, url, ssl=<context>)` (or `ssl=False`) succeeds and does not raise `TypeError: unexpected keyword argument 'ssl'`. The `ssl` option is forwarded into the underlying client request handling path.
2. Every previously supported call pattern of `aiohttp.client.request`... | model |
aio-libs__aiohttp-10300 | aio-libs/aiohttp | 10,300 | python | 3d3fba23cf543f7185658f52097f21093bf64c7c | scope | Scope clarity first. This ticket is intentionally narrow.
In scope:
- Modifying the top-level `request` function in `aiohttp/client.py` so that it accepts keyword arguments drawn from `_RequestOptions` (such as `ssl`) and forwards them into the underlying client request handling path.
- Adding or updating unit tests i... | model |
aiscript-dev__aiscript-811 | aiscript-dev/aiscript | 811 | ts | e02c847a2ad7abd107af5c126ef03ab41a22591d | narrative | ## Problem
AiScript runtime errors currently report only the location where the failure occurred, which makes it painful to diagnose problems in scripts that compose multiple functions. When an exception is raised deep inside a nested call chain, the user has to manually reconstruct the calling sequence from source.
... | model |
aiscript-dev__aiscript-811 | aiscript-dev/aiscript | 811 | ts | e02c847a2ad7abd107af5c126ef03ab41a22591d | acceptance | ## Done when (acceptance criteria)
1. A runtime error thrown from inside a user-defined function has a `message` that ends with a call stack listing every frame from the failing call upward to `<root>`, in order from innermost to outermost.
2. Each frame in the call stack renders as `at <name> (Line <n>, Column <m>)` ... | model |
aiscript-dev__aiscript-811 | aiscript-dev/aiscript | 811 | ts | e02c847a2ad7abd107af5c126ef03ab41a22591d | scope | ## Scope
### In scope
- Attaching a call stack to AiScript interpreter runtime errors so that the message text includes the chain of calls and source locations leading to the failure.
- The stack is appended to the error `message` string. Each frame is prefixed with `at <name>` and includes `Line <n>, Column <m>` whe... | model |
aiscript-dev__aiscript-825 | aiscript-dev/aiscript | 825 | ts | d4496b30d48fc9be483c1b8ebea0e1c230eb4ff5 | narrative | ## Summary
The parser for `match` expressions currently allows a `case` clause to flow directly into a `default` clause without any intervening separator (semicolon, comma, or newline). Every other clause-to-clause transition already requires a separator, so the `case`-to-`default` path is an outlier. This task closes... | model |
aiscript-dev__aiscript-825 | aiscript-dev/aiscript | 825 | ts | d4496b30d48fc9be483c1b8ebea0e1c230eb4ff5 | acceptance | ## Done when...
1. Given a `match` expression with two consecutive `case` clauses and no separator between them, the parser produces a syntax error.
2. Given a `match` expression with a `case` clause immediately followed by a `default` clause and no separator, the parser produces a syntax error.
3. Given a `match` exp... | model |
aiscript-dev__aiscript-825 | aiscript-dev/aiscript | 825 | ts | d4496b30d48fc9be483c1b8ebea0e1c230eb4ff5 | scope | ## Scope
This task covers exactly the following:
- The parser-level handling of the `match` expression, specifically the clause grammar inside the `{ ... }` block.
- Refactoring `parseMatch` into three functions aligned with the ABNF: `parseMatch`, `parseMatchCase`, and `parseDefaultCase`.
- Introducing a shared `par... | model |
aiscript-dev__aiscript-827 | aiscript-dev/aiscript | 827 | ts | d4496b30d48fc9be483c1b8ebea0e1c230eb4ff5 | narrative | ## Intent
The aiscript parser/semantic analyzer must be adjusted so that single-statement bodies belonging to `if`, `match`, and `for` constructs open their own lexical scope. The one exception is when the lone statement is already a block expression — in that case the existing block scope is reused so the construct d... | model |
aiscript-dev__aiscript-827 | aiscript-dev/aiscript | 827 | ts | d4496b30d48fc9be483c1b8ebea0e1c230eb4ff5 | acceptance | ## Done When...
1. `if true let a = 1` followed by a reference to `a` outside the `if` produces the runtime error `No such variable 'a' in scope '<root>' (Line N, Column M)`.
2. `if false null elif true let a = 1` produces the same error when `a` is referenced outside the `elif`.
3. `if false null else let a = 1` prod... | model |
aiscript-dev__aiscript-827 | aiscript-dev/aiscript | 827 | ts | d4496b30d48fc9be483c1b8ebea0e1c230eb4ff5 | scope | ## In Scope
- The aiscript parser and/or semantic analyzer scope-creation logic for the body of an `if` expression (then clause, elif clause, else clause).
- The body of a `match` expression (case clause, default clause).
- The body of a `for` statement.
- Specifically: when each of those bodies consists of exactly on... | model |
aiscript-dev__aiscript-870 | aiscript-dev/aiscript | 870 | ts | cdb1b17146320683b5e865a1b6a24b9c0ded0a9b | narrative | ## Task: Accept additional type annotations in AiScript
### Intent
Bring the AiScript parser and type checker in line with the documented type-annotation syntax by supporting generic parameters, union types, the `error` type, and the `never` type inside function declarations, function bodies, function-type expressions... | model |
aiscript-dev__aiscript-870 | aiscript-dev/aiscript | 870 | ts | cdb1b17146320683b5e865a1b6a24b9c0ded0a9b | acceptance | ## Task: Accept generics, union, `error`, and `never` type annotations in AiScript
### Done when...
1. `fn<T>(x: T): T { return x }` parses and type-checks successfully.
2. `Fn<T> :: (x: T) => T` parses and type-checks successfully.
3. `let v: number | string = 0` parses and type-checks successfully.
4. An annotation ... | model |
aiscript-dev__aiscript-870 | aiscript-dev/aiscript | 870 | ts | cdb1b17146320683b5e865a1b6a24b9c0ded0a9b | scope | ## Task: Accept documented type-annotation forms in AiScript
### In scope
This task covers the parser and type-checker changes needed to accept four annotation forms that the AiScript documentation already advertises:
1. Generic type parameters in function declarations, with references to those parameters allowed fro... | model |
alangpierce__sucrase-788 | alangpierce/sucrase | 788 | ts | 365fccc84dd0eb52910f86bdb6078bae47d3bb27 | narrative | ## Feature: Add a "preserve" value to the jsxRuntime option
**Intent.** Introduce a new mode for the JSX transform in sucrase by accepting the string `"preserve"` as a valid value for the existing `jsxRuntime` configuration option. The `jsx` transform must continue to be registered as an active transform alongside thi... | model |
alangpierce__sucrase-788 | alangpierce/sucrase | 788 | ts | 365fccc84dd0eb52910f86bdb6078bae47d3bb27 | acceptance | ## Add `"preserve"` mode to the JSX transform
### Done when...
1. The `jsxRuntime` option accepts the string `"preserve"` as a valid value without raising an error from the option validator.
2. When the `jsx` transform is enabled and `jsxRuntime` is `"preserve"`, JSX syntax (e.g., `<div>foo</div>`) appears verbatim i... | model |
alangpierce__sucrase-788 | alangpierce/sucrase | 788 | ts | 365fccc84dd0eb52910f86bdb6078bae47d3bb27 | scope | ## Add `"preserve"` mode to the JSX transform
### In scope
This task covers three things:
1. **Option surface.** Accept `"preserve"` as a new valid value for the existing `jsxRuntime` option. The `jsx` transform must continue to be registered as an active transform even when `"preserve"` is in use, because the parse... | model |
alangpierce__sucrase-811 | alangpierce/sucrase | 811 | ts | a924d84d606c43cd900c3ffd793c25bd9f44ce7f | narrative | ## Add a `keepUnusedImports` compile option to Sucrase
### Intent
Introduce a new compile-time flag, `keepUnusedImports`, that prevents Sucrase from eliding import and export statements (and their named bindings) during the TypeScript, Flow, and imports transforms. This flag addresses the elision aspect of TypeScript... | model |
alangpierce__sucrase-811 | alangpierce/sucrase | 811 | ts | a924d84d606c43cd900c3ffd793c25bd9f44ce7f | acceptance | ## `keepUnusedImports` option for Sucrase — acceptance-led spec
### Done when...
1. **`keepUnusedImports: true` with the `typescript` transform** preserves every `import` and `export` statement — and every named binding within them — verbatim in the output, even when zero references to the imported names remain in th... | model |
alangpierce__sucrase-811 | alangpierce/sucrase | 811 | ts | a924d84d606c43cd900c3ffd793c25bd9f44ce7f | scope | ## Scope: add `keepUnusedImports` to Sucrase
### In scope
1. **Sucrase core transforms** — `typescript`, `imports`, `flow`. Each must honor a new `keepUnusedImports` option that disables automatic import/export elision and named-binding elision.
2. **Public transform-options API** — the option is a new boolean field ... | model |
aleph-alpha__ts-rs-389 | Aleph-Alpha/ts-rs | 389 | rust | ee2240db942b7e32b185e5028e1d3cb91f7fae4b | narrative | ### Intent
Rework the `TS` derive macro's parser for `#[serde(...)]` attributes so that encountering an unrecognized inner setting no longer discards the entire attribute. Instead, the parser should skip ahead to the next comma and keep parsing the remaining (supported) settings. Additionally, separate the handling of... | model |
Statement Portfolio
Redrafted problem statements for 2,094 feature-implementation instances from
nebius/SWE-rebench-V2,
as used in the statement-portfolio experiments of our paper. Each instance has three
redrafts of the same required change:
variant |
Style |
|---|---|
narrative |
Explains the requested change and its intended behavior in prose |
acceptance |
Leads with the criteria that a completed implementation must meet |
scope |
States what the change should include and exclude |
Redrafts were written by a multi-agent pipeline (two drafters, cross-critique, arbiter) that reads the public pull-request record (linked issues, review discussion, PR description, commit messages). The reference patch and hidden tests were never shown to the drafting models.
Fields
| Field | Description |
|---|---|
instance_id |
SWE-rebench-V2 instance id (join key) |
repo, pr_number |
Source repository and pull request |
language |
Repository language |
base_commit |
Repository snapshot the task starts from |
variant |
narrative, acceptance, or scope |
problem_statement |
The redrafted statement, exactly as given to the agent |
generation |
How the statement was produced |
6,282 rows (2,094 instances × 3 variants).
Reconstructing executable tasks
This release contains statements only. Repository snapshot, reference patch, test patch,
FAIL_TO_PASS/PASS_TO_PASS tests, and environment configuration come from
SWE-rebench-V2 at revision 475dd5e8703bb5fb22dd3c60b5d038b019eba1e0; join on
instance_id.
License and citation
Released under CC-BY-4.0, following SWE-rebench-V2. Please cite SWE-rebench-V2 and our paper when using this data.
- Downloads last month
- 24