Environment:

• pandas==0.24.1
• -e git+ssh://git@github.com/KuguHome/influxdb-python.git@afcfd25b21523d84a7d1088eff2abb4d08de7647#egg=influxdb which mirrors the current master branch of influxdb-python
• python=3.6.7

Reproduction of bug:
bug-reproduce.py

import pandas as pd
from influxdb import DataFrameClient

dfi = DataFrameClient(database='testdb', )
dfi.create_database('testdb')

# create a test dataframe that reproduces the bug
buggy_df = pd.DataFrame(
	dict(
		first=[1, None, None, 8, 9],
		second=[2, None, None, None, 10],
                third=[3, 4.1, None, None, 11],
		first_tag=["one", None, None, "eight", None],
		second_tag=["two", None, None, None, None],
                third_tag=["three", "four", None, None, None],
                comment=[
                    "All columns filled",
                    "First two of three empty",
                    "All empty",
                    "Last two of three empty",
                    "Empty tags with values",
                ]
	),
	index=pd.date_range(
		start=pd.to_datetime('2018-01-01'),
		end=pd.to_datetime('2018-10-01'),
		periods=5,
	)
)
print("bug inducing df=\n",buggy_df)
print("\nbuggy line protocol output\n",
      "\n" + "\n".join(
		dfi._convert_dataframe_to_lines(
			buggy_df.iloc[:,:-1],
			'buggy-measurements',
			tag_columns=["first_tag", "second_tag","third_tag"],
		)
	)
)

$ python bug-reproduce.py
bug inducing df=
                      first  second  third first_tag second_tag third_tag                   comment
2018-01-01 00:00:00    1.0     2.0    3.0       one        two     three        All columns filled
2018-03-10 06:00:00    NaN     NaN    4.1      None       None      four  First two of three empty
2018-05-17 12:00:00    NaN     NaN    NaN      None       None      None                 All empty
2018-07-24 18:00:00    8.0     NaN    NaN     eight       None      None   Last two of three empty
2018-10-01 00:00:00    9.0    10.0   11.0      None       None      None    Empty tags with values

buggy line protocol output
 
buggy-measurements,first_tag=one,second_tag=two,third_tag=three first=1.0,second=2.0,third=3.0 1514764800000000000
buggy-measurements,third_tag=four ,third=4.1 1520661600000000000
buggy-measurements,first_tag=eight first=8.0 1532455200000000000
buggy-measurements first=9.0,second=10.0,third=11.0 1538352000000000000

Outlook:
Although tags and fields are handled the same way, the presence of a leading comma works for the tags. The following code is where the values are handled :

Notice the difference to how tags are handled, in that the None columns are filtered out by value before conversion to line protocol. The values, however, are handled by the column index i.e field_df[field_df.columns[1:]] = ',' + field_df[ field_df.columns[1:]]

Fix and refactor:

1. [FIX] I would have proposed that the fields are handled exactly the same way as the tags. However, the tag operation uses apply (which is slower), whereas the field operation uses vectorized operation. So the performant fix here would be to strip the leading , from fields where appropriate.
   KuguHome@b4ceab9
   Output:

   python bug-reproduce.py
   bug inducing df=
                         first  second  third first_tag second_tag third_tag                   comment
   2018-01-01 00:00:00    1.0     2.0    3.0       one        two     three        All columns filled
   2018-03-10 06:00:00    NaN     NaN    4.1      None       None      four  First two of three empty
   2018-05-17 12:00:00    NaN     NaN    NaN      None       None      None                 All empty
   2018-07-24 18:00:00    8.0     NaN    NaN     eight       None      None   Last two of three empty
   2018-10-01 00:00:00    9.0    10.0   11.0      None       None      None    Empty tags with values
   
   buggy line protocol output
    
   buggy-measurements,first_tag=one,second_tag=two,third_tag=three first=1.0,second=2.0,third=3.0 1514764800000000000
   buggy-measurements,third_tag=four third=4.1 1520661600000000000
   buggy-measurements,first_tag=eight first=8.0 1532455200000000000
   buggy-measurements first=9.0,second=10.0,third=11.0 1538352000000000000

2. [REFACTOR] Since the logic for handling tags and values are exactly the same, these two chunks should be refactored into a utility function called __lineify_tag_field_df, because DRY.
   The leading , needs to be taken out from the tags output and needs to be put back when constructing the line again.
   KuguHome@366e771
   Output:

   python bug-reproduce.py           
   bug inducing df=
                         first  second  third first_tag second_tag third_tag                   comment
   2018-01-01 00:00:00    1.0     2.0    3.0       one        two     three        All columns filled
   2018-03-10 06:00:00    NaN     NaN    4.1      None       None      four  First two of three empty
   2018-05-17 12:00:00    NaN     NaN    NaN      None       None      None                 All empty
   2018-07-24 18:00:00    8.0     NaN    NaN     eight       None      None   Last two of three empty
   2018-10-01 00:00:00    9.0    10.0   11.0      None       None      None    Empty tags with values
   
   buggy line protocol output
    
   buggy-measurements,first_tag=one,second_tag=two,third_tag=three first=1.0,second=2.0,third=3.0 1514764800000000000
   buggy-measurements,third_tag=four third=4.1 1520661600000000000
   buggy-measurements,first_tag=eight first=8.0 1532455200000000000
   buggy-measurements first=9.0,second=10.0,third=11.0 1538352000000000000

Pull request at #694