[Data] Improve appearance of repr(dataset) #59631

myandpr · 2025-12-23T15:44:50Z

Description

improve the UX by making this look like polars when using Ray Dataset.

test demo

>>> import numpy as np
>>> import ray
>>> from ray.data import DataContext

ray.init(include_dashboard=False, ignore_reinit_error=True)
>>> 
>>> ray.init(include_dashboard=False, ignore_reinit_error=True)
2025-12-25 03:38:11,358 INFO worker.py:2010 -- Started a local Ray instance.
/Users/xxx/work/community/ray/python/ray/_private/worker.py:2049: FutureWarning: Tip: In future versions of Ray, Ray will no longer override accelerator visible devices env var if num_gpus=0 or num_gpus=None (default). To enable this behavior and turn off this error message, set RAY_ACCEL_ENV_VAR_OVERRIDE_ON_ZERO=0
  warnings.warn(
RayContext(dashboard_url='', python_version='3.10.19', ray_version='3.0.0.dev0', ray_commit='{{RAY_COMMIT_SHA}}')

# Configure truncation: show 4 rows total (1 from the head, the rest from the tail)
# and display up to 5 columns (2 from the head, 2 from the tail, plus an ellipsis).
>>> ctx = DataContext.get_current()
>>> ctx.dataset_repr_max_rows = 4        # Display a total of 4 rows
>>> ctx.dataset_repr_head_rows = 1       # To display 1 row from the head and the remaining from the tail
>>> ctx.dataset_repr_max_columns = 5     # Show 5 columns in total, with middle columns truncated (represented by ellipsis ...).
>>> ctx.dataset_repr_head_columns = 2    # Display the first 2 columns at the head, and the remaining columns at the tail

# Create a demo dataset with 10 rows and 6 columns.
>>> items = [
...     {f"col{i}": i + row for i in range(6)}
...     for row in range(10)
... ]
>>> ds = ray.data.from_items(items)

>>> 
>>> print("Before materialization (schema preview only):")
Before materialization (schema preview only):
>>> print(ds)                    
shape: (10, 6)
╭───────┬───────┬─────┬───────┬───────╮
│ col0  ┆ col1  ┆ …   ┆ col4  ┆ col5  │
│ ---   ┆ ---   ┆ --- ┆ ---   ┆ ---   │
│ int64 ┆ int64 ┆ …   ┆ int64 ┆ int64 │
╰───────┴───────┴─────┴───────┴───────╯
(Showing 0 of 10 rows)
(Showing 4 of 6 columns)
>>> print("\nAfter materialization (shows head/tail rows):")

After materialization (shows head/tail rows):
>>> print(ds.materialize())      # To display a head/tail summary of 4 rows with truncated columns/ellipsis in a tabular format
shape: (10, 6)
╭───────┬───────┬─────┬───────┬───────╮
│ col0  ┆ col1  ┆ …   ┆ col4  ┆ col5  │
│ ---   ┆ ---   ┆ --- ┆ ---   ┆ ---   │
│ int64 ┆ int64 ┆ …   ┆ int64 ┆ int64 │
╞═══════╪═══════╪═════╪═══════╪═══════╡
│ 0     ┆ 1     ┆ …   ┆ 4     ┆ 5     │
│ …     ┆ …     ┆ …   ┆ …     ┆ …     │
│ 7     ┆ 8     ┆ …   ┆ 11    ┆ 12    │
│ 8     ┆ 9     ┆ …   ┆ 12    ┆ 13    │
│ 9     ┆ 10    ┆ …   ┆ 13    ┆ 14    │
╰───────┴───────┴─────┴───────┴───────╯
(Showing 4 of 10 rows)
(Showing 4 of 6 columns)

Related issues

Link related issues: #59482

Fixes #59482

gemini-code-assist

Code Review

This pull request introduces a new tabular representation for ray.data.Dataset, similar to polars, which significantly improves the user experience for inspecting datasets. The implementation is well-structured, with a fallback mechanism to the old plan-based representation to ensure robustness. The changes include new helper functions for building the ASCII table, and the tests have been updated accordingly to cover both the new and old repr behaviors.

I've found one area for performance improvement in the logic for collecting tail rows for the representation, which can be made more efficient. Overall, this is a great enhancement to Ray Data.

python/ray/data/dataset.py

myandpr · 2025-12-23T15:56:00Z

Hi @bveeramani — I put together this PR to implement the Polars-style repr requested in #59482. When you have a moment, could you take a quick look and let me know if this matches what you had in mind? Happy to adjust or follow up on any feedback.

python/ray/data/dataset.py

bveeramani · 2025-12-23T18:38:04Z

python/ray/data/dataset.py

+    column_widths: List[int],
+) -> List[str]:
+    lines: List[str] = []
+    top = _render_border("┌", "┬", "┐", "─", column_widths)


Nit: Could we use rounded corners? I think people find them more aesthetically pleasing

I have used rounded corners, and it looks like more aesthetically. Thanks for your suggestion

bveeramani · 2025-12-23T18:39:04Z

python/ray/data/dataset.py


 logger = logging.getLogger(__name__)

+_DATASET_REPR_MAX_ROWS = 10


What happens if you have 1000 columns? How does polars or other DataFrame libraries handle this?

@myandpr I would suggest reading through: https://github.com/pola-rs/polars/blob/ac0b7519/crates/polars-core/src/fmt.rs .

They handle truncation at both column and row level, and both should be configurable from the user's perspective.

We should test rendering of large vectors, strings etc. in repr to ensure that truncation works as expected. Each dtype should have its own config

@goutamvenkat-anyscale , Thanks for the pointers! Dataset.repr now uses the polars-style table, truncation knobs (rows/columns/value per dtype) are exposed via DataContext, and tests cover row/column truncation plus long string/list/ndarray cases.

I'm not entirely sure if this matches what you were envisioning, so please let me know if anything still needs adjusting.

bveeramani · 2025-12-23T18:39:26Z

@myandpr Nice!! This is great 🔥

@owenowenisme could you review the implementation pls?

python/ray/data/dataset.py

python/ray/data/tests/test_consumption.py

python/ray/data/dataset.py

owenowenisme · 2025-12-25T07:10:37Z

python/ray/data/dataset.py

    return block.to_arrow()
+
+
+@dataclass


Could we move these implementations into a separate module? dataset.py is already ~7,000 lines, and continuing to append to it will hurt maintainability.

Thanks for your review, and I have moved these implementations into a separate dataset_repr.py file.

owenowenisme · 2025-12-25T07:56:32Z

python/ray/data/dataset.py

+    return head_rows, tail_rows, show_gap
+
+
+def _determine_preview_row_targets(


Can we add some comments for this?

python/ray/data/_internal/dataset_repr.py

owenowenisme · 2025-12-26T05:30:50Z

@myandpr The test is failing, could you fix that?

python/ray/data/_internal/dataset_repr.py

myandpr · 2025-12-30T16:09:14Z

@myandpr The test is failing, could you fix that?

Hi @owenowenisme ,I’ve fixed the failing tests—could you take another look when you have time? Thanks a lot!

bveeramani

ty! Just left some comments.

Ping me whenever you're ready for another round of feedback, and I can give you a fast review

bveeramani · 2025-12-31T05:39:17Z

doc/source/data/loading-data.rst

+            :options: +ELLIPSIS

-            MaterializedDataset(
-               num_blocks=3,
-               num_rows=3,
-               schema={food: string, price: double}
-            )
+            shape: (3, 2)
+            ...
+            (Showing 3 of 3 rows)


(Here and elsewhere) What happens if we show the actual repr rather than ellipsis?

I'm sure it's a pain to update, but I think it'll be a lot more useful to our readers. For example, in the screenshot below, it's not clear what the schema or data look like.

Updated the doc examples in the files touched by this PR to show the full actual repr instead of .../ELLIPSIS (including loading-data.rst and quickstart.rst).

bveeramani · 2025-12-31T05:40:32Z

python/ray/data/context.py

+DEFAULT_DATASET_REPR_MAX_ROWS = env_integer("RAY_DATASET_REPR_MAX_ROWS", 10)
+DEFAULT_DATASET_REPR_HEAD_ROWS = env_integer("RAY_DATASET_REPR_HEAD_ROWS", 5)
+DEFAULT_DATASET_REPR_MAX_COLUMNS = env_integer("RAY_DATASET_REPR_MAX_COLUMNS", 10)
+DEFAULT_DATASET_REPR_HEAD_COLUMNS = env_integer("RAY_DATASET_REPR_HEAD_COLUMNS", 5)
+DEFAULT_DATASET_REPR_MAX_COLUMN_WIDTH = env_integer(
+    "RAY_DATASET_REPR_MAX_COLUMN_WIDTH", 40
+)
+DEFAULT_DATASET_REPR_MAX_STRING_LENGTH = env_integer(
+    "RAY_DATASET_REPR_MAX_STRING_LENGTH", 40
+)
+DEFAULT_DATASET_REPR_MAX_BYTES_LENGTH = env_integer(
+    "RAY_DATASET_REPR_MAX_BYTES_LENGTH", 40
+)
+DEFAULT_DATASET_REPR_MAX_COLLECTION_ITEMS = env_integer(
+    "RAY_DATASET_REPR_MAX_COLLECTION_ITEMS", 5
+)
+DEFAULT_DATASET_REPR_MAX_TENSOR_ELEMENTS = env_integer(
+    "RAY_DATASET_REPR_MAX_TENSOR_ELEMENTS", 8
+)
+


What happens if we don't expose these configurations? Is it important that users are able to configure these?

If it is important that we expose these to the user, I think it'd be cleaner if we encapsulate this within a single dataclass to avoid flooding the top-level DataContext with several repr-specific parameters

IMO I don't think it's super important to expose these knobs, but would love to get your take

Thanks for the feedback! I agree these knobs don’t need to be user‑exposed right now. I removed the dataset_repr_* fields from DataContext and kept the defaults internal to the repr implementation (and dropped the config-based tests accordingly).

bveeramani · 2025-12-31T05:40:50Z

python/ray/data/dataset.py

+        except Exception:
+            logger.debug("Falling back to plan-based Dataset repr.", exc_info=True)
+            return self._plan.get_plan_as_string(self.__class__)


When would this fail?

This is a safety net for failures in the tabular repr path (e.g., schema fetch / BlockAccessor / pyarrow issues or formatting errors).
If anything in that path throws, we fall back to the plan-based repr so repr(ds) never crashes user workflows.

I don't think we should do a bare except because it prevents us from removing the old code path, and it'll also mask bugs that we would want to hear about and fix

bveeramani · 2025-12-31T05:41:37Z

python/ray/data/dataset.py

+        if schema is None or not isinstance(schema, Schema):
+            return self._plan.get_plan_as_string(self.__class__)


Is the only code path where we use get_plan_as_string?

get_plan_as_string is used as a fallback in repr (when tabular rendering fails or schema is missing) and also when the dataset has no columns in
_build_dataset_ascii_repr. Since it’s a low‑level detail, I removed tests that asserted it directly.

bveeramani · 2025-12-31T05:43:11Z

python/ray/data/tests/test_consumption.py



-def test_dataset_repr(ray_start_regular_shared):
+def test_dataset_plan_string(ray_start_regular_shared):


get_plan_as_string is a low-level implementation detail, and I don't think we need to test it.

I think we should either test the actual repr(ds), or if itds redundant with the tests below, I think we can jsut nuke this test

Agreed. I removed the tests that directly assert get_plan_as_string (both test_dataset_plan_string and test_dataset_plan_as_string) to avoid coupling to low‑level plan formatting.

bveeramani · 2025-12-31T05:47:01Z

python/ray/data/tests/test_consumption.py

+    assert "Dataset isn't materialized" in text
+    assert "shape: (5, 1)" in text
+    assert "│ id" in text


Also, should we explicitly assert that the repr looks like the expected string? I think it might be easier to read and catch more errors, but I guess the tradeoff is that it'd make the test pretty brittle

bveeramani · 2025-12-31T05:47:22Z

python/ray/data/tests/test_consumption.py

+    ctx.dataset_repr_max_rows = 4
+    ctx.dataset_repr_head_rows = 1


Related to my comment above -- if we don't need to expose these configuration, I don't think we need these tests

bveeramani · 2025-12-31T05:50:37Z

python/ray/data/tests/test_groupby_e2e.py

    ]
-    assert "MapBatches(func)" in ds.__repr__()
+    plan_repr = ds._plan.get_plan_as_string(type(ds))
+    assert "MapBatches(func)" in plan_repr



With this and test_map -- would it make sense to assert against Dataset.explain() instead?

Yes—updated both test_groupby_e2e and test_map to assert against Dataset.explain() output instead of repr / get_plan_as_string.

bveeramani · 2025-12-31T05:51:31Z

python/ray/train/base_trainer.py

+def _truncate_repr_if_needed(representation: str) -> str:
+    """Ensure repr strings remain shorter than MAX_REPR_LENGTH."""
+    if len(representation) < MAX_REPR_LENGTH:
+        return representation
+
+    ellipsis = "..."
+    closing = ""
+    if representation:
+        closing = representation[-1]
+        representation = representation[:-1]
+
+    max_body_len = MAX_REPR_LENGTH - len(ellipsis) - len(closing) - 1
+    if max_body_len < 0:
+        max_body_len = 0
+
+    truncated_body = representation[:max_body_len]
+    return f"{truncated_body}{ellipsis}{closing}"
+
+


What happens if we don't include this change?

Without this, repr(trainer) can exceed MAX_REPR_LENGTH when non‑default args (like datasets/config) are large. There’s an existing test (test_base_trainer.py::test_repr) that asserts len(repr(trainer)) < MAX_REPR_LENGTH, so this truncation keeps that invariant.

Oh, I see. I think the problem is that for Train V1, it uses the dataset repr in the Trainer repr, and that looks pretty bad with the table format:

"<DummyTrainer datasets={'train': shape: (3, 1)\n╭───────╮\n│ item │\n│ --- │\n│ int64 │\n╞═══════╡\n│ 1 │\n│ 2 ...>"

Rather than truncate the repr, I think we should special-case datasets in the BaseTrainer repr and just do dataset._plan.get_plan_as_string(...). It's not great because it breaks abstraction barriers and uses a brittle low-level API, but I think it might be the best option for now because it's simple and we can't break Train V1

bveeramani · 2025-12-31T05:54:44Z

python/ray/data/_internal/dataset_repr.py

+    is_gap: bool = False
+
+
+def _prepare_columns_for_repr(


As I read the functions in this module, it wasn't obvious to me what each of the functions did without reading the implementation. Would you mind adding docstrings summarizing what each function does (not just repeating the funtion name), and describing the input and output types?

added docstring

myandpr · 2026-01-09T16:31:42Z

Conflicts fixed.

bveeramani

ty! Left a bunch of comments.

Some high-level feedback:

I had difficulty reading dataset_repr.py. To make the module easier to read, I think we should remove unnecessary indirection (e.g., _DatasetReprConfig ) and cut scope (e.g., remove special cases that might not be necessary for this initial implementation)
Could you split out the pre-requisite test refactors into different PRs? We can land the changes faster that way

bveeramani · 2026-01-10T03:23:51Z

doc/source/data/loading-data.rst

                "food": ["spam", "ham", "eggs"],
                "price": [9.34, 5.37, 0.94]
            })
+            df["food"] = df["food"].astype("string")


What does the output look like if we exclude this? I think readers of the documentation might feel confused as to why we include this in the code snippet

bveeramani · 2026-01-12T06:48:42Z

python/ray/data/context.py


 DEFAULT_ENABLE_GET_OBJECT_LOCATIONS_FOR_METRICS = False

-


Nit: Unintended diff?

bveeramani · 2026-01-12T06:50:33Z

python/ray/data/_internal/dataset_repr.py

+@dataclass
+class _DatasetReprConfig:
+    max_rows: int
+    head_rows: int
+    max_columns: int
+    head_columns: int
+    max_column_width: int
+    max_string_length: int
+    max_bytes_length: int
+    max_collection_items: int
+    max_tensor_elements: int
+
+


Now that we've removed these configurations from the DataContext, this dataclass is a layer of indirection. Could you remove it, and also all of the dead logic? I think it'd really simplify some of this code

For example, once we remove _DatasetReprConfig , we can remove this snippet:

Before

max_columns = config.max_columns if max_columns is None or max_columns <= 0: max_columns = len(columns) max_columns = max(1, max_columns) if num_columns <= max_columns: ...

After

if num_columns <= _DATASET_REPR_MAX_COLUMNS: ...

bveeramani · 2026-01-12T06:51:24Z

python/ray/data/_internal/dataset_repr.py

+_DATASET_REPR_MAX_STRING_LENGTH = 40
+_DATASET_REPR_MAX_BYTES_LENGTH = 40


Do we need to define lengths for strings and bytes? Can we just truncate based on the column width?

bveeramani · 2026-01-12T06:52:34Z

python/ray/data/_internal/dataset_repr.py

+_DATASET_REPR_ELLIPSIS = "…"
+_DATASET_REPR_PREVIEW_MAX_DEPTH = 2
+_DATASET_REPR_MAX_ROWS = 10
+_DATASET_REPR_HEAD_ROWS = 5
+_DATASET_REPR_MAX_COLUMNS = 10
+_DATASET_REPR_HEAD_COLUMNS = 5


Could you add comments describing what each of these constants represent? I felt confused by _DATASET_REPR_PREVIEW_MAX_DEPTH and _DATASET_REPR_HEAD_COLUMNS in particular

bveeramani · 2026-01-12T07:30:40Z

python/ray/data/dataset.py

+        except Exception:
+            logger.debug("Falling back to plan-based Dataset repr.", exc_info=True)
+            return self._plan.get_plan_as_string(self.__class__)


I don't think we should do a bare except because it prevents us from removing the old code path, and it'll also mask bugs that we would want to hear about and fix

bveeramani · 2026-01-12T07:37:34Z

python/ray/data/tests/test_stats.py

+    plan_str = ds._plan.get_plan_as_string(type(ds))
+    assert plan_str == (
        "MapBatches(<lambda>)\n"
        "+- Dataset(name=test_ds, num_rows=100, schema={id: int64})"
    )


I don't think this assertion will make a lot of sense after this PR, because it's testing against an extremely low-level API that's mostly obsolete. I think the intention of this assertion is to make sure we include the dataset name in the repr and that the dataset name is correct.

If the user has specified a name, should we include that somewhere in the repr?

For the assertion here and later in this function, I think we should just test that the name is in the repr, like assert "test_ds" in repr(ds)

bveeramani · 2026-01-12T07:38:59Z

python/ray/data/tests/test_tensor.py

-    assert str(ds) == (
+    plan_str = ds._plan.get_plan_as_string(type(ds))
+    assert plan_str == (


Could you update the assertions in this file to explicitly check the count and scheme rather than test against get_plan_as_string?

bveeramani · 2026-01-12T07:55:38Z

python/ray/train/base_trainer.py

+def _truncate_repr_if_needed(representation: str) -> str:
+    """Ensure repr strings remain shorter than MAX_REPR_LENGTH."""
+    if len(representation) < MAX_REPR_LENGTH:
+        return representation
+
+    ellipsis = "..."
+    closing = ""
+    if representation:
+        closing = representation[-1]
+        representation = representation[:-1]
+
+    max_body_len = MAX_REPR_LENGTH - len(ellipsis) - len(closing) - 1
+    if max_body_len < 0:
+        max_body_len = 0
+
+    truncated_body = representation[:max_body_len]
+    return f"{truncated_body}{ellipsis}{closing}"
+
+


Oh, I see. I think the problem is that for Train V1, it uses the dataset repr in the Trainer repr, and that looks pretty bad with the table format:

"<DummyTrainer datasets={'train': shape: (3, 1)\n╭───────╮\n│ item │\n│ --- │\n│ int64 │\n╞═══════╡\n│ 1 │\n│ 2 ...>"

Rather than truncate the repr, I think we should special-case datasets in the BaseTrainer repr and just do dataset._plan.get_plan_as_string(...). It's not great because it breaks abstraction barriers and uses a brittle low-level API, but I think it might be the best option for now because it's simple and we can't break Train V1

bveeramani · 2026-01-12T07:59:31Z

python/ray/data/_internal/dataset_repr.py

+def _build_dataset_ascii_repr(
+    dataset: "Dataset",
+    schema: "Schema",
+    is_materialized: bool,
+    config: _DatasetReprConfig,
+) -> str:


This isn't easily testable because it requires a Ray cluster and can't be truly unit tested.

To make this easy to quickly test (and so that we can easily test all of the edge and special cases), could you split this into two functions:

One higher-level functions that accepts a dataset and maybe some other parameters as input

A lower-level function that just accepts a schema, row count, dataset name, and head and tail dict[str, Any] rows?

And then we can write tests for most of the cases with (2), put them in data/tests/unit, and they'll run immediately

cursor · 2026-01-13T21:24:08Z

python/ray/data/_internal/dataset_repr.py

+def _format_value(value: Any) -> str:
+    if isinstance(value, np.generic):
+        value = value.item()
+    return str(value)


Newlines in cell values break table rendering

Medium Severity

The _format_value function uses str(value) directly, which for numpy arrays produces strings containing embedded newlines (e.g., '[[1. 1.]\n [1. 1.]]' for a 2x2 array). Since this 20-character string is shorter than _DATASET_REPR_MAX_COLUMN_WIDTH (40), it won't be truncated by _truncate_to_cell_width. The newline character is then passed to _render_row, where ljust() doesn't handle it, causing the table border structure to break across multiple lines and corrupting the ASCII table output.

Signed-off-by: yaommen <[email protected]>

myandpr requested a review from a team as a code owner December 23, 2025 15:44

gemini-code-assist bot reviewed Dec 23, 2025

View reviewed changes

python/ray/data/dataset.py Outdated Show resolved Hide resolved

cursor bot reviewed Dec 23, 2025

View reviewed changes

python/ray/data/dataset.py Outdated Show resolved Hide resolved

bveeramani reviewed Dec 23, 2025

View reviewed changes

ray-gardener bot added data Ray Data-related issues community-contribution Contributed by the community labels Dec 23, 2025

cursor bot reviewed Dec 24, 2025

View reviewed changes

python/ray/data/dataset.py Outdated Show resolved Hide resolved

cursor bot reviewed Dec 25, 2025

View reviewed changes

python/ray/data/tests/test_consumption.py Outdated Show resolved Hide resolved

myandpr requested a review from a team as a code owner December 25, 2025 06:47

cursor bot reviewed Dec 25, 2025

View reviewed changes

python/ray/data/dataset.py Outdated Show resolved Hide resolved

owenowenisme reviewed Dec 25, 2025

View reviewed changes

myandpr requested review from bveeramani, goutamvenkat-anyscale and owenowenisme December 25, 2025 16:11

cursor bot reviewed Dec 25, 2025

View reviewed changes

python/ray/data/_internal/dataset_repr.py Outdated Show resolved Hide resolved

python/ray/data/_internal/dataset_repr.py Outdated Show resolved Hide resolved

bveeramani added the go add ONLY when ready to merge, run all tests label Dec 26, 2025

myandpr force-pushed the improve-repr-dataset branch from 5d5ed05 to b26992f Compare December 29, 2025 12:47

cursor bot reviewed Dec 29, 2025

View reviewed changes

python/ray/data/_internal/dataset_repr.py Outdated Show resolved Hide resolved

python/ray/data/_internal/dataset_repr.py Show resolved Hide resolved

bveeramani reviewed Dec 31, 2025

View reviewed changes

myandpr force-pushed the improve-repr-dataset branch 2 times, most recently from e32fa05 to 6af81f2 Compare January 9, 2026 16:29

myandpr requested a review from bveeramani January 10, 2026 03:35

owenowenisme approved these changes Jan 10, 2026

View reviewed changes

owenowenisme mentioned this pull request Jan 12, 2026

Improve dataset repr with polars-style table formatting #59748

Closed

bveeramani reviewed Jan 12, 2026

View reviewed changes

myandpr force-pushed the improve-repr-dataset branch from 6af81f2 to a7565c5 Compare January 13, 2026 21:12

cursor bot reviewed Jan 13, 2026

View reviewed changes

[Data] Improve appearance of repr(dataset)

0b8fc4d

Signed-off-by: yaommen <[email protected]>

myandpr force-pushed the improve-repr-dataset branch from a7565c5 to 0b8fc4d Compare January 14, 2026 07:43


		logger = logging.getLogger(__name__)

		_DATASET_REPR_MAX_ROWS = 10

		return head_rows, tail_rows, show_gap


		def _determine_preview_row_targets(

		if schema is None or not isinstance(schema, Schema):
		return self._plan.get_plan_as_string(self.__class__)



		def test_dataset_repr(ray_start_regular_shared):
		def test_dataset_plan_string(ray_start_regular_shared):

		_DATASET_REPR_MAX_STRING_LENGTH = 40
		_DATASET_REPR_MAX_BYTES_LENGTH = 40

[Data] Improve appearance of repr(dataset) #59631

Are you sure you want to change the base?

[Data] Improve appearance of repr(dataset) #59631

Conversation

myandpr commented Dec 23, 2025 • edited by bveeramani Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Description

test demo

Related issues

Uh oh!

gemini-code-assist bot left a comment

Choose a reason for hiding this comment

Code Review

Uh oh!

Uh oh!

myandpr commented Dec 23, 2025

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

bveeramani commented Dec 23, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

owenowenisme commented Dec 26, 2025

Uh oh!

Uh oh!

Uh oh!

myandpr commented Dec 30, 2025

Uh oh!

bveeramani left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

myandpr commented Dec 23, 2025 •

edited by bveeramani

Loading

bveeramani commented Dec 23, 2025 •

edited

Loading

bveeramani left a comment •

edited

Loading