Help me love Python and yaml (tongue in check)

[DanPartelly]

1. Pro:
   • Netlab is written in Python
   • Netlab uses yaml
   • The network automation world uses Python instead of a decent language, even a demi-decent one
   • I have itches to scratch in a language I don't want to learn
2. Against
   Do you really wanna read a Dune-sized novel on Python Butlerian Jihad? And on why you shouldn't describe trees with white space?

Smile and have a nice day. Thank you all for you contributions to Netlab.

[jbemmel]

This may provide a perspective: https://github.com/exergy-connect/netlab.js

The viewer at https://exergy-connect.github.io/netlab.js/ converts YAML into YANG-validated JSON. It is so strict it will make Ivan look like a <blank>...

[DanPartelly]

Would love to know more. Maybe we should put together a list of Box gotchas.

Look at 67d5edb, in the wip ACL branch. Particularly at this construct:
global_acls = topology.routing.get("acl", {})

That single line was enough to break hundreds of transformation tests. But ... really ? Crash early is better than this crap.

[ipspace]

Would love to know more. Maybe we should put together a list of Box gotchas.

Look at 67d5edb, in the wip ACL branch. Particularly at this construct: global_acls = topology.routing.get("acl", {})

That single line was enough to break hundreds of transformation tests. But ... really ? Crash early is better than this crap.

Yeah, that's the Box "feature" I like and other people hate (see #3513). Without automatic creation of nested dictionaries a lot of code would look way more awkward than it does now because we'd have to check "do we have the parent dictionary already" every time we'd create a nested element. But like any sharp tool, it cuts both ways.

The drawback is what you discovered -- if you're not careful, you're creating stuff that shouldn't be there. Anyway, in your case, topology.get('routing.acl',{}) should do the trick.

We have a similar dilemma in Jinja2 templates -- do we allow "a.b.c.d is defined" (which works now after some heavy arm-twisting) or would we prefer to have "a is defined and a.b is defined and a.b.c is defined and a.b.c.d is defined" every time.

[DanPartelly]

I understand you like it. It's convenient. Once you know it, it's easy to avoid. But it breaks any algebra, for it breaks the identity operation. That makes it a big piece of banta poodo in my book. I have rarely seen shit like this in 39 years.

I'm sure you know that I'm not proposing changing anything. Humans adapt and go on, and rewriting big chunks of software that performs well is something I would never recommend or go for. But allow me to express my disgust :P

[ipspace]

I understand you like it. It's convenient. Once you know it, it's easy to avoid.

Exactly. The only question is "how do we make people aware of which closet doors not to open because a skeleton might fall out"

But it breaks any algebra, for it breaks the identity operation. That makes it a big piece of banta poodo in my book. I have rarely seen shit like this in 39 years.

Think of it as a function call with side effects, because a.b.c is really a series of get function calls. If I understood it correctly, even with Python objects, a.b.c turns into a series of getattr calls. And yes, that's also disgusting.

But allow me to express my disgust :P

While we're doing that, can I express my utter disgust for (standard Python) def x(a: list = [ 1,2,3 ]) which results in modifying the default value whenever you touch "a" in an x() call.

[DanPartelly]

Think of it as a function call with side effects, because a.b.c is really a series of get function calls. If I understood it correctly, even with Python objects, a.b.c turns into a series of getattr calls. And yes, that's also disgusting.

Thinking of it in terms of side effects is OK. But it's not really my point.

If identity is broken, associativity is broken. Good luck doing map-reduce on something like that. Composition may or may not work, depending of the case. Not that we need any of it, but you get the idea. If you look at early STL designs, the guys (probably Alex Stepanov) who did that were extremely careful to preserve elementary algebras. Despite side effects everywhere.

Anyway, rant over. Back to enjoying Netlab =)

[ipspace]

Smile and have a nice day.

As it's been well-known for at least 40 years, real programmers can write FORTRAN programs in any language (https://www.pbm.com/~lindahl/real.programmers.html ;))

No tool is perfect, and some tools are better than others, but we can get a job done anyway ;)

Thank you all for you contributions to Netlab.

Couldn't agree more. Thank you all!