invio / invio.hashing Goto Github PK

Background
As part of our migration to v1.0.4 of the .NET SDK, I set the DebugType to full and updated the CI scripts to explicitly reference XML-based *.csproj files due to the removal of global.json and project.json. The DebugType really should be portable for .NET Core when we pack the code (though it needs to be full for when we run opencover on it), and we should add a *.sln to manage our *.csproj files.

Task

• Update the repository to have a Invio.Hashing.sln solution at the root of the repository that references the source and test projects.
• Update Invio.Hashing.Tests.csproj such that <IsPackable>false</IsPackable> is specified.
• Restore set-debug-type.ps1, but with it updating the DebugType via the XML-based Invio.Hashing.csproj instead of JSON-based project.json.

If there are a list of values that you are generating a hash code for, you may want to calculate the hash code of those values using a specific IEqualityComparer implementation rather than use the object's default GetHashCode() implementation.

For example, you may want to use a case-insensitive string comparison for objects of type string, but the normal hash code for all other objects.

Our open source libraries are currently fractured between two versions of the SDK:
v1.0.1 and v1.0.0-preview2-003121.

Let's get these standardized on the latest version supported by our CI tools (presumably v1.0.4).

Often I find myself in a situation where I have a list of objects that may contain strings. If there is a string, I want to do a case-insensitive equality/hash code comparison for each string found. Otherwise, use the default equals/hash code generation and comparison. I think this could best be solved by an IfStringComparer implementation that accepts objects instead of strings, and uses a predefined string comparison syntax for each string found, and the normal GetHashCode() implementation on all other objects (and a predefined constant for null - including null strings).

If we want to hash a list and maintain order, we need to implement our own hashing algorithm every time. If we want to hash a set and ignore order, we need to implement our own hashing algorithm every time.

I just wrote a bug in another repository because I incorrectly implemented my own hashing algorithm that we could solve once with a generic solution here. Let's do that instead.

NOTE: I don't think we need to update the main From implementation to do this by default for objects it can coerce as IEnumerable<T>, IList<T>, ISet<T>, et cetera. This is because I do not want to slow down the implementation with branches. I also believe it would make the hashing implementation too opinionated on what to do with a given IEnumerable (specifically: whether or not to respect order).

Travis CI has not yet been updated to use .NET Core v1.1. Code coverage and deployment is performed via appveyor, so it is not a blocking issue. However, we will not be testing to make sure the library works on Linux / Mac OS X with Mono or .NET Core.

The underlying issue from Travis CI is here.