Add SingleUsePassword

[adrian-salajan]

One might want to use a password just once to initialise something, after that the password should not stay in memory.

[codecov]

Codecov Report

Merging #388 into master will not change coverage.
The diff coverage is 100.00%.

@@            Coverage Diff            @@
##            master      #388   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files           15        16    +1     
  Lines          388       394    +6     
  Branches        15        21    +6     
=========================================
+ Hits           388       394    +6     

Impacted Files	Coverage Δ
...ules/core/src/main/scala/ciris/ConfigDecoder.scala	100.00% <100.00%> (ø)
.../core/src/main/scala/ciris/SingleUsePassword.scala	100.00% <100.00%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 1fef96f...8e24290. Read the comment docs.

[adrian-salajan]

I'm not sure which implementation is the better/correct one
SingleUsePassword[F[_]] private (private val password: PasswordProtection) - current PR
vs
SingleUsePassword[F[_]] private (private val password: Resource[F,PasswordProtection]) - first commit in this PR

[vlovgr]

Hey, thanks for raising this.

You might want to have a look at SafePassword as reference.

[adrian-salajan]

Hey, thanks for raising this.

You might want to have a look at SafePassword as reference.

Hi, I saw it, actually it was what triggered this attempt :)

1. i guess it is better to use PasswordProtection as a core type
2. it might be safer to not use String anywhere, for construction or returning the password, to avoid any JVM string interning stuff
3. i suspect my initial design of the class SingleUsePassword[F[_]] private (private val password: Resource[F,PasswordProtection]) might be safer since it is a description of a computation so it avoids any storing of the password until the computation is run

[vlovgr]

@adrian-salajan Sorry for leaving this for so long. But I'd like to get this merged (in some form) for the next release candidate. I've also opened #457 to showcase an alternative implementation to the one in this pull request.

i guess it is better to use PasswordProtection as a core type

I don't see it providing any benefit. It uses Arrays.fill(chars, ' ') which is what I've opted for in #457 as well.

it might be safer to not use String anywhere, for construction or returning the password, to avoid any JVM string interning stuff

I agree it makes sense to use Array[Char] everywhere for security, even though it's mutable. I've opted for this in #457.

[vlovgr]

Closing in favour of #457.