Comments (9)
fjetter
commented on June 24, 2024
1
Thanks for chasing this down! I agre this should raise, I opened #11174 to address this
from dask.
fjetter
commented on June 24, 2024
what pyarrow version did you use to create the dataset and which one are you using to read the dataset?
from dask.
dbalabka
commented on June 24, 2024
@fjetter, thanks for a quick reply. We installed pyarrow 14.0.2.
The important observation is that TypeError error disappears if I take only part of the dataset as follows:
ddf.loc[:100000]However, disabling the dask-expr still leads to an error:
ValueError: Length of values (0) does not match length of index (100001)
(I've added a note to the description)
from dask.
dbalabka
commented on June 24, 2024
@fjetter , in case of TypeError: cannot concatenate object of type '<class 'tuple'>'; only Series and DataFrame objs are valid I'm getting the following values here:
https://github.com/dask/dask/blob/main/dask/dataframe/backends.py#L688
I've added the debug code on all nodes as follows:
from dask.distributed import print
print(dfs3[0], dfs3[1], dfs3, join)
out = pd.concat(dfs3, join=join, sort=False)
dfs3[0] - is DataFrame
dfs3[1] - is tuple with the following value:
('repartition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc', 4)
dfs3 is a list that looks like as follows:
[ .. .[48081 rows x 4 columns], ('repartition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc', 4), ('repartition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc', 5), ('repartition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc', 6), ('repartition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc', 7), ('repa
rtition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc', 8), ('repartition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc', 9), ('repartition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc', 10), ('repartition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc', 11), ('repartition-split-100000000-fe2ba6
93bab4c1021c4766fe26e0d5dc', 12), ('repartition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc', 13), ('repartition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc', 14), ('repartition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc', 15), ('repartition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc',
16), ('repartition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc', 17), ('repartition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc', 18), ('repartition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc', 19), ('repartition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc', 20), ('repartition-split-10
0000000-fe2ba693bab4c1021c4766fe26e0d5dc', 21), ('repartition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc', 22), ('repartition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc', 23), ('repartition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc', 24), ('repartition-split-100000000-fe2ba693bab4c1021c47
66fe26e0d5dc', 25), ('repartition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc', 26), ('repartition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc', 27), ('repartition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc', 28), ('repartition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc', 29), ('repart
ition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc', 30), ('repartition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc', 31), ('repartition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc', 32), ('repartition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc', 33), ('repartition-split-100000000-fe2ba6
93bab4c1021c4766fe26e0d5dc', 34), ('repartition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc', 35), ('repartition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc', 36), ('repartition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc', 37), ('repartition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc',
38), ('repartition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc', 39), ('repartition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc', 40), ('repartition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc', 41), ('repartition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc', 42), ('repartition-split-10
0000000-fe2ba693bab4c1021c4766fe26e0d5dc', 43), ('repartition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc', 44), ('repartition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc', 45), ('repartition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc', 46), ('repartition-split-100000000-fe2ba693bab4c1021c47
66fe26e0d5dc', 47), ('repartition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc', 48), ('repartition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc', 49), ('repartition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc', 50), ('repartition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc', 51), ('repart
ition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc', 52), ('repartition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc', 53), ('repartition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc', 54), ('repartition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc', 55), ('repartition-split-100000000-fe2ba6
93bab4c1021c4766fe26e0d5dc', 56), ('repartition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc', 57), ('repartition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc', 58), ('repartition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc', 59), ('repartition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc', 60), ('repartition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc', 61), ('repartition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc', 62), ('repartition-split-100000000-fe2ba693bab4c1021c4766fe26e0d5dc', 63)]
from dask.
dbalabka
commented on June 24, 2024
@fjetter , in case of ValueError: Length of values (0) does not match length of index (12295809) I'm getting the following values here:
https://github.com/dask/dask/blob/main/dask/dataframe/backends.py#L523-L525
I've added the debug code on all nodes as follows:
@hash_object_dispatch.register((pd.DataFrame, pd.Series, pd.Index))
def hash_object_pandas(
obj, index=True, encoding="utf8", hash_key=None, categorize=True
):
from dask.distributed import print
print(obj)
print(index)
print(encoding)
print(hash_key)
print(categorize)
return pd.util.hash_pandas_object(
obj, index=index, encoding=encoding, hash_key=hash_key, categorize=categorize
)
Here is an output:
Empty DataFrame
Columns: []
Index: [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, ...]
[12295809 rows x 0 columns]
False
utf8
None
True
from dask.
dbalabka
commented on June 24, 2024
@fjetter , the problem is that I'm shuffling using the existing column. Dask should provide a proper error message in this case. I've added preproduction steps. Please see:
- Use any parquet data
- Try to shuffle using a non-existing column and export the data:
ddf = (dd
.read_parquet('gs://.../....parquet')
.shuffle(on='does_not_exist', npartitions=64)
.repartition(partition_size='100MB')
.to_parquet('data_test.parquet'))
from dask.
dbalabka
commented on June 24, 2024
@fjetter , the problem isn't strictly related to dask-expr. The different problem appears even when dask-expr is off using dask.config.set({"dataframe.query-planning": False}) as described in the ticket's description.
from dask.
fjetter
commented on June 24, 2024
indeed, I transferred the issue to dask/dask
from dask.
dbalabka
commented on June 24, 2024
@fjetter , after spending some time testing the solution, I've found that there are two separate problems:
Problem 1
The following problem still appears when the dask-expr is enabled, and I don't have a reproduction step yet.
TypeError: cannot concatenate object of type '<class 'tuple'>'; only Series and DataFrame objs are valid
Please see more details here: #11160 (comment)
Problem 2
The following problem is gone and appears only when dask-expr is disabled and incorrect on=<column> provided for shuffle() method:
ValueError: Length of values (0) does not match length of index (100001)
Please see #11160 (comment)
from dask.
Related Issues (20)
- Concat with unknown divisions raises TypeError HOT 1
- Dask 2024.5.1 removed `.attrs` HOT 11
- Dask 2024.5.1 raises exception when `.compute()` is called on a categorical column HOT 3
- a tutorial for distributed text deduplication HOT 5
- Large memory use when loading file with np.memmap in recent dask versions HOT 8
- `dask/dataframe/tests/test_indexing.py::test_getitem_integer_slice` failing with nightly `pandas`
- dask.dataframe import error for Python 3.12.3 HOT 3
- 'SeriesGroupBy' object has no attribute 'nunique_approx' HOT 6
- Categorical column information incorrectly copied over when using series to create new dataframe resulting in a broken dataframe
- calling repartition on ddf with timeseries index after resample causes ValueError: left side of old and new divisions are different
- P2P rechunking of ERA-5 from spatial to temporal dimension is failing hard HOT 15
- Improve documentation for `dd.from_map(...)` HOT 1
- AssertionError: DataFrame are different with dask 2024.5.1 and python 3.12 HOT 3
- `test_quantile` flaky
- Shuffle not raising exception when `on` does not exist HOT 1
- [FEA] Add official mechanism to check if query-planning is enabled in ``dask.dataframe`` HOT 3
- UnboundLocalError in test_dt_accessor when dd._dask_expr_enabled is False HOT 2
- Error with the default tokenizer. HOT 4
- Cannot bind async delayed
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from dask.