[WIP] configurable encoding + optimizing json access

[isoos]

No description provided.

[codecov-commenter]

Codecov Report

Patch coverage: 82.85% and project coverage change: -0.05% ⚠️

Comparison is base (7bfe932) 86.51% compared to head (715caa7) 86.47%.

❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #132      +/-   ##
==========================================
- Coverage   86.51%   86.47%   -0.05%     
==========================================
  Files          29       30       +1     
  Lines        2492     2521      +29     
==========================================
+ Hits         2156     2180      +24     
- Misses        336      341       +5     

Files Changed	Coverage Δ
lib/src/text_codec.dart	80.70% <0.00%> (ø)
lib/src/v3/types.dart	67.64% <0.00%> (-9.78%)	⬇️
lib/src/logical_replication_messages.dart	71.42% <72.41%> (ø)
lib/src/v3/connection.dart	85.10% <75.00%> (-0.21%)	⬇️
lib/src/v3/protocol.dart	85.71% <88.88%> (ø)
lib/src/binary_codec.dart	97.54% <94.11%> (+0.02%)	⬆️
lib/postgres_v3_experimental.dart	51.16% <100.00%> (ø)
lib/src/client_encoding.dart	100.00% <100.00%> (ø)
lib/src/connection.dart	93.59% <100.00%> (ø)
lib/src/message_window.dart	100.00% <100.00%> (ø)
... and 2 more

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[isoos]

/cc @simolus3 @insinfo

I was trying to adapt the optimization from #130 on the encoding PR from @insinfo, but it seems to be a huge drag on the codebase, and it will increase the boilerplate required for many objects (incl. server and client messages that were not all covered on the encoding PR).

I'm open for suggestions, however, if there is no better option, I shall complete this PR with the missing parts (utf8_encoded_string, and the configuration settings on both connection)...

[simolus3]

I don't see a much better way to do this either. Maybe we could put the message serialization / deserialization logic into a single class and turn the existing classes into fairly dumb data classes. That might reduce some of the duplicate parameters in the message layer. But I'm not sure if this actually simplifies things or just moves the problem around.

[isoos]

Closing this as I think a refactoring of the message classes is required before we can introduce encoding efficiently.