Fix `get_route_to` or fix tests

[dbarrosop]

The get_route_to method is a tricky one as the protocol argument is not thoroughly tested and might lead to differences in terms of support across different drivers. For example, a driver could support static and bgp while another might support only static.

Proposed solutions:

1. get_route_to only lives in napalm-base and calls for get_route_to_static, get_route_to_bgp, etc... that are individual implementations in the drivers. Would be nice to have an IPv4/IPv6 flag, btw.
2. We ignore this issue and we just fix the tests so there is no "protocol" sent to get_route_to and control tests/behaviour with mocked data as described here: Test case for static routes (v4+v6) #233 (comment)

Pinging some people that might be interested on this discussion: @napalm-automation/council @manuel-domke @bewing

[mirceaulinic]

I have commented under #233 (comment) -- I don't see (1) as a sane solution. I tend to consider (2) more suitable.

[dbarrosop]

(answering to your comment here instead of on the other thread)

remove an useful functionality

Nobody wants to remove anything, just move it around.

why not just implement ping_vrf, ping_ttl

Easy to answer that and I will just refer to the code:

https://github.com/napalm-automation/napalm-eos/blob/develop/napalm_eos/eos.py#L1054

Protocol is a switch to go to a branch of the code that contains ~100 lines of code. Adding a new protocol implies branching again and executing 100 other different lines of code.

In any case, please, read point 1 again. Nobody mentioned removing anything, jut moving get_route_to to napalm-base and have it call different methods for each individual protocol. Something like:

# This is not working code, just for reference
def get_route_to(self, protocol='', ...):
    if protocol in ["direct", "bgp", "ospf", "static"...]:
        method = getattr(self, "get_route_to_{}".format(protocol))
        return method(...)
    elif protocol == "":
       # code to call all of getters and ignore notimplementederrors
   else:
      # fail as protocol is incorrect

And then each driver just has to implement get_route_to_static, get_route_to_bgp, etc. which serves as implementation and documentation. You can refer to the docs and see that get_route_ospf is not implemented and thus it won't work. Right now it's just ignored as you suggested.

[mirceaulinic]

Thanks for expanding on this @dbarrosop - indeed, I misread the comments above, apologies.

Makes sense to me, however it may take longer to get this done :)

[dbarrosop]

Makes sense to me, however it may take longer to get this done :)

Agreed, but hopefully things will be easier in the future so we may save time in the long run.

[madhurabhogate]

I had an issue if the route installed was through ibgp then it used to error out as -

line 1071, in get_route_to
remote_as = int(as_path.split()[-1])
IndexError: list index out of range