• Switch parsing to Superpower (from Sprache)
• Fix utf8 parsing
• Interpolated variables parsing

• Use ConcurrentDictionary for FakeEnvVars for thread safety

• Retain interpolation when NoEnvVars() is set (setEnvVars == false)
• NoClobber works correctly with ConfigurationProvider

• Parse unquoted values slightly differently to properly parse comments, rejecting quote chars, allow leading and trailing whitespace in interpolated values, do not allow values to start with # anymore, give better errors on invalid quoted values, etc
• Rename ToDictionary to ToDotEnvDictionary for clarity, and add options for behavior
• Get rid of the warnings about vulnerabilities in System.Net.Http 4.3.0 (which was unused)

Revert the change from Directory.GetCurrentDirectory() to AppContext,BaseDirectory as it broke existing applications

DO NOT USE -- the change in directory did in fact break behavior for at least one existing application. Move to 2.5.0

• Replace the default path from Directory.GetCurrentDirectory() with just AppContext.BaseDirectory -- which should do the same thing, except make things better for some cases, per #69
• Add the newly contributed ConfigurationBuilder integration

Adds a new EnvReader helper class #65

Allow dash and period in env var identifiers

Properly handle dollar signs in values

https://www.nuget.org/packages/DotNetEnv/2.1.0

Technically this includes 2 breaking changes, but it is highly unlikely anyone is using them, and the changes are clear improvements.

• Adds Env.LoadMulti for loading multiple .env files on top of each other
• Removes Env.Load for list of lines -- was only used for tests here and highly discouraged for deployed use as it effectively defines hardcoded env vars in code. If you need to reproduce this functionality, use LoadContents with "\n"s between lines in a single string.
• Unquoted values can now include inline whitespace -- but still allow for comments at the end, even while allowing inline hashes in a word (to support k8s configmap from env)

Again, note that this changes the behavior of unquoted values in a noticeable way -- but before this would have been a parse error, so this simply makes it more flexible than it used to be.