Currently, JSONLexerRange eagerly skips whitespace for performance reasons, but everything else is evaluated lazily. The problem with eager WS skipping is that it makes the lexer unusable in a request/response scenario where a requests consists of a single JSONValue. Skipping the whitespace at the end would block on the connection until the next request arrives, which may never happen if the previous one never gets processed.
This should either be made configurable, or the fully lazy version needs to be brought up to the same speed as the eager version.
Panke
commented
3 years ago
Currently, JSONLexerRange eagerly skips whitespace for performance reasons, but everything else is evaluated lazily. The problem with eager WS skipping is that it makes the lexer unusable in a request/response scenario where a requests consists of a single JSONValue. Skipping the whitespace at the end would block on the connection until the next request arrives, which may never happen if the previous one never gets processed.
This should either be made configurable, or the fully lazy version needs to be brought up to the same speed as the eager version.