Read parquet file error

Hi,

I'm reading parquet files and facing some problems. For comparison the file was read with python using fastparquet with no errors.

The file contains 74 columns and already the output of parquetinfo shows that there is an issue with some of the columns, since the length of the variables doesn't match:

FileSize: 66748042

NumRowGroups: 1

RowGroupHeights: 233382

VariableNames: [1×31 string]

VariableTypes: [1×31 string]

VariableCompression: [1×74 string]

When reading the data with Matlab the following error occurs:

tbl = parquetread(fname)

Error using matlab.io.parquet.internal.makeParquetException>makeUnsupportedParquetTypeException (line 26)

To assign to or create a variable in a table, the number of rows must match the height of the table.

Error in matlab.io.parquet.internal.makeParquetException (line 10)

e = makeUnsupportedParquetTypeException(e, filename);

Error in parquetread (line 128)

e = makeParquetException(e, filename);

I've tried to access the data via the column name directly using the column names provided by parquetinfo. This works for some variables but for other variables it states that the variable is not a subset of the variablenames:

tbl = parquetread(fname ,'SelectedVariableNames','variableX');

Error using parquetread (line 124)

'SelectedVariableNames' value must be a unique subset of 'variable1, ....'

Using the list of variables in this error message delivers again the first error message

tbl = parquetread(fname ,'SelectedVariableNames','variable1');

me =

MException with properties:

identifier: 'MATLAB:table:RowDimensionMismatch'

message: 'To assign to or create a variable in a table, the number of rows must match the height of the table.'

cause: {}

stack: [3×1 struct]

Correction: []

The I tried to access the data via the column names coming from python. However, this gievs the error that the variable is not part of the variable list (see second error above)

In Python some of the column names contain a period like "name.name1". Can this be an issue in Matlab?

Any ideas on that?

Thanks a lot.

Christian

Answer 1

[Yongjian Feng]

matlab can call python script. Use python to read and then pass the data to matlab. Or use python to convert the data into CSV or json first and then use matlab to read.

[Christian]

Thanks for the answer. That's my current workaround. However, I thought there might/should be a pure Matlab way to do it.

[tim che]

could you please tell how to read file using python? I have this problem

[Yongjian Feng]

This looks like error shown in matlab, right? If so, maybe you need to make sure the python script can run successfully outside matlab first. Then figure out how to run it from matlab.