Compute hash of OpenApi document

[MaggieKimani1]

Fixes #799

[baywet]

I don't think we should be using the GetHashCode override for multiple reasons:

• it's not its intended use, especially if caller want to store the hash somewhere for later comparison
• it can yield equality between unequal objects
• it will vary depending on chip instruction set (x86, 64, arm, arm64)
• it requires also overriding the Equals method

I'd instead recommend using a proper hashing algorithm (configurable?) like sha256 or up.

see the documentation of get hash code.
https://docs.microsoft.com/en-us/dotnet/api/system.object.gethashcode?view=net-6.0#remarks

[baywet]

I think you're supposed to select different prime numbers.

[MaggieKimani1]

I don't think we should be using the GetHashCode override for multiple reasons:

• it's not its intended use, especially if caller want to store the hash somewhere for later comparison
• it can yield equality between unequal objects
• it will vary depending on chip instruction set (x86, 64, arm, arm64)
• it requires also overriding the Equals method

I'd instead recommend using a proper hashing algorithm (configurable?) like sha256 or up.

see the documentation of get hash code. https://docs.microsoft.com/en-us/dotnet/api/system.object.gethashcode?view=net-6.0#remarks

I leaned more towards overriding GetHashCode() and tried getting the hash value for each of the property values to avoid collision as I believed the goal was to perform a simple object lookup as opposed to encryption which comes with its own set of complications:

• You have to convert the OpenApi document object into a byte array for hash computation
• We have to mark all object model types, OpenApiAny and Expression types as either Serializable or NonSerialized for the hashing to take place..

[baywet]

le object lookup as opposed to encryption which comes with its own set of complications:

• You have to convert the OpenApi document object into a byte array for hash computation
• We have to mark all object model types, OpenApiAny and Expression types as Serializable for the hashing to take place..

Nitpicking: hashing and encryption are two completely different things :) I'm not suggesting to rely on any encryption algorithm, simply on standard hashing.

The nice additional thing with using industry standard hashing is that if we document the process correctly, other libs (from other languages) could interop with it and generate the exact same hash for the same document.

Serialize and convert: well that's the beauty of it, we already have the infrastructure to serialize the document to JSON/YAML, let's use that to produce a terse representation (so white-spacing doesn't impact the result), convert to binary, hash, and voilà! You don't even need to implement changes on the whole object model anymore, just on the document or root object.

What do you think? Also @darrelmiller, any objection with this approach?

[baywet]

Thanks for making the changes, a few minor comments.
We should also make sure the serialization is terse when we hash

[baywet]

thanks for making the changes, here are a few final remarks