Remove requests requirement, and use urllib built-in modules

[mwtoews]

Hi Mike, I'm experiencing issues with the web_service module, e.g.

from hilltoppy import web_service as ws

base_url = 'https://data.hbrc.govt.nz/Envirodata'
hts = 'ContinuousArchive.hts'
site = 'Well.16772 Ngatarawa Rd'
measurement = 'Elevation Above Sea Level[Recorder Water Level]'
ts_df = ws.get_data(base_url, hts, site, measurement)
ts = ts_df.reset_index().set_index("DateTime")["Value"]

raises "ChunkedEncodingError: ("Connection broken: ConnectionResetError(104, 'Connection reset by peer')", ConnectionResetError(104, 'Connection reset by peer'))".

This error is the same on Windows and Linux using recentish packages from conda-forge.

Is this #29 and/or #30?

A fix (for me at least) is to simply replace Python Requests with the urllib built-in modules. It doesn't seem that the high-level features of Requests is warranted by this module, as it's pretty basic encode and retrieve data from a stream.

I've mostly rewritten build_url to build a dict of data that gets encoded once. The order of the dict should be preserved, even for Python 3.6, since most folks use CPython. (The insertion order is guaranteed from Python 3.7). The previous build_url encoded parts, but not all. This will send perhaps different URLs. So, check carefully with your use-cases. Also interested to know how this change behaves with large data requests. I've only been using this module since yesterday, so I'm actually a novice with this package.

[mullenkamp]

Thanks for that. I do wonder what's going wrong with requests...
Let's give it a try, and I'll definitely test the larger datasets.
I've been meaning to rewrite some of the code in the build_url function. I wrote that code fairly early on in my Python coding career ;)