This PR is a preliminary re-implementation of the appendContent function which uses out-file with the -append parameter instead of echo w/ a redirect. Using echo w/ a redirect will typically result in a unicode-type encoding which uses 16-bit wchar_t internally resulting in twice as much base64 encoded data needing to be decoded. By using out-file with a utf8 encoding, this should halve the base64 data that is written to the file which "might" make things a little faster. I think we can use 7-bit ascii due to base64, but to be safe I chose utf8 because it's universal.
However, I believe the same amount of data is being transferred over a socket, but during the decoding process, less data needs to be processed. Also, to execute an i/o redirect the command has to be executed, have its output captured, a handle needs to be created for the file, data needs to be appended (a seek), and then the handle has to be closed. Using out-file keeps all functionality within the same implementation and doesn't need to process any of the extra features of write-output which supports different logging classes and other special formatting (which can have some implicit performance issues).
This is based on the first suggestion I made in PR #6. It hasn't been tested too well, but should have the exact same semantics as the original implementation. I have not comprehensively measured its performance. :-/
This PR is a preliminary re-implementation of the
appendContentfunction which usesout-filewith the-appendparameter instead ofechow/ a redirect. Usingechow/ a redirect will typically result in a unicode-type encoding which uses 16-bit wchar_t internally resulting in twice as much base64 encoded data needing to be decoded. By usingout-filewith a utf8 encoding, this should halve the base64 data that is written to the file which "might" make things a little faster. I think we can use 7-bit ascii due to base64, but to be safe I chose utf8 because it's universal.However, I believe the same amount of data is being transferred over a socket, but during the decoding process, less data needs to be processed. Also, to execute an i/o redirect the command has to be executed, have its output captured, a handle needs to be created for the file, data needs to be appended (a seek), and then the handle has to be closed. Using
out-filekeeps all functionality within the same implementation and doesn't need to process any of the extra features ofwrite-outputwhich supports different logging classes and other special formatting (which can have some implicit performance issues).This is based on the first suggestion I made in PR #6. It hasn't been tested too well, but should have the exact same semantics as the original implementation. I have not comprehensively measured its performance. :-/