pandas_to_eland, dot case columns naming raises a mapper_parsing_exception

[OdelinCharron]

Hello,

I got a request error when using eland.pandas_to_eland.
This is happening when using this.type.of.column.name. instead of camelCaseNames` (for example).

Eland version: 7.10.1b1
EK stack 7.10.0 (docker)

How to reproduce:

df = pandas.DataFrame({
    "a":[1,2,3,4,5],
    "b":["a","b","c","d","e"],
    "b.c":[1,2,3,4,5]
})

	a	b	b.c
0	1	a	1
1	2	b	2
2	3	c	3
3	4	d	4
4	5	e	5

eland.pandas_to_eland(
    pd_df = df,
    es_client=es,
    es_dest_index="test_eland",
    es_if_exists="replace",
    es_refresh=True,
)

returns:

RequestError: RequestError(400, 'mapper_parsing_exception', 'Failed to parse mapping [_doc]: mapper [b] cannot be changed from type [keyword] to [ObjectMapper]')

There are very quick fix to solve this issue like df.columns = [_c.replace(".","_") for _c in df.columns] but that was not a behavior I expected.

Feel free it close it directly if it's only a lack of culture on my side.

Thanks !

[stevedodson]

@OdelinCharron - thanks for the feedback. In Elasticsearch dots are special characters in field names (elastic/elasticsearch#15951). In your example, field b needs to be both a list and an object (b.c implies b: { c: [1,2,3,4,5] }) so Elasticsearch rejects it.

Closing this for now, but happy for more feedback.