insert_dataframe() need dataframe's column order exactly same as clickhouse table's column order

[zsmeijin]

Describe the bug

In BlockOutputStream class, I found following code in write method:

for i, (col_name, col_type) in enumerate(block.columns_with_types):
    write_binary_str(col_name, self.fout)
    write_binary_str(col_type, self.fout)

    if n_columns:
        try:
            items = block.get_column_by_index(i)
        except IndexError:
            raise ValueError('Different rows length')
        write_column(self.context, col_name, col_type, items,
                     self.fout, types_check=block.types_check)

it seems column data in block is got by index, rather than column name. so if dataframe's column order is not same as clickhouse table's column order, the data type can go wrong.

By now reindex dataframe's column order before insert is enough to solve this problem. I'm not sure if it is a design to improve insert performance, but at least this additional requirement should be documented

To Reproduce

use insert_dataframe() method to insert a dataframe who's column order is not same as clickhouse table's column order

Expected behavior

dataframe’s column and clickhouse table's column can be matched by name, rather than index

Versions

Version of package with the problem. = 0.2.1
Python version. = 3.8.9

[xzkostyan]

Fix with arbitrary columns order was merged into master branch.