Eliminates redundant objects creations in NSData base64 decoding #2773
Conversation
Sources/Foundation/NSData.swift
Outdated
| @@ -670,7 +671,7 @@ open class NSData : NSObject, NSCopying, NSMutableCopying, NSSecureCoding { | |||
| - parameter options: Options for handling invalid input | |||
| - returns: The decoded bytes. | |||
| */ | |||
| private static func base64DecodeBytes(_ bytes: [UInt8], options: Base64DecodingOptions = []) -> [UInt8]? { | |||
| private static func base64DecodeBytes(_ bytes: UnsafeRawBufferPointer, options: Base64DecodingOptions = []) -> [UInt8]? { | |||
There was a problem hiding this comment.
Rather than accepting an UnsafeRawBufferPointer it would be safer to change the signature to :
private static func base64DecodeBytes<T: Collection>(_ bytes: T, options: Base64DecodingOptions = []) -> [UInt8]? where T.Element == UInt8 {
Then it could be called using:
NSData.base64DecodeBytes(base64String.utf8, ...)
and
NSData.base64DecodeBytes(base64Data, ...)
This has a 2nd benefit of avoiding the use of .utf8CString which may have to reallocate if it needs more space since it appends a NUL to the end of the buffer. This also eliminates the use of withUnsafeBytes in the 2nd initialiser.
What do you think?
|
@swift-ci test linux |
4 similar comments
|
@swift-ci test linux |
|
@swift-ci test linux |
|
@swift-ci test linux |
|
@swift-ci test linux |
|
This is fine although could you just squash it down into a single commit please. I do have some decoding optimisations to add but I can rebase them on top of these ones. |
…to NSData base64 decoding.
valeriyvan
force-pushed
the
RedundantZeroingNSData
branch
from
0d55e8d to
b070a51
Compare
|
Squashed |
|
@swift-ci please test |
1 similar comment
|
@swift-ci please test |
|
@swift-ci test and merge |
|
@swift-ci please test and merge |
No description provided.