Create Perspective Table from empty Arrow Table

[0x26res]

Feature Request

Description of Problem:

I'm hosting arrow tables on a perspective server.

I get an error when I try to host empty tables. By empty table, I mean table with zero rows, but with columns and a well defined schema.

import pyarrow as pa
import perspective

perspective_server = perspective.Server()
client = perspective_server.new_local_client()

table = pa.table(
    {
        "col1": [1, 2, 3],
        "col2": ["abc", "foo", "bar"],
    }
)
empty_table = table.schema.empty_table()

client.table(table, name="table")
client.table(empty_table, name="table_empty_bad") # Throws an error

As a work around I have to explicitly pass the schema when the table is empty:

client.table({"col1": "integer", "col2": "string"}, name="table_empty_from_schema")

Potential Solutions:

I think it should be a simple change around this line:

perspective/cpp/perspective/src/cpp/arrow_loader.cpp

Line 94 in 452bfb0

auto status3 = arrow::Table::FromRecordBatches(batches);

As a side question, could you make available a util function to convert from pyarrow.Schema to the perspective schema (eg: {"col1": "integer", "col2": "string"}).