taepper commented on code in PR #50945:
URL: https://github.com/apache/arrow/pull/50945#discussion_r3978794379
##########
cpp/src/arrow/json/chunker.cc:
##########
@@ -165,6 +91,78 @@ class ParsingBoundaryFinder : public BoundaryFinder {
int64_t* out_pos, int64_t* num_found) override {
return Status::NotImplemented("ParsingBoundaryFinder::FindNth");
}
+
+ private:
+ simdjson::ondemand::parser parser_;
+ // A persistent buffer to keep padded contents for simdjson.
+ // This should be more efficient than allocating a new padded_string
everytime.
+ std::string buffer_;
+
+ simdjson::padded_string_view GetPaddedStringView(std::string_view partial,
+ std::string_view block =
{}) {
+ // Adjust buffer size without copying old contents.
+ buffer_.clear();
+ buffer_.reserve(partial.size() + block.size() +
simdjson::SIMDJSON_PADDING);
+ buffer_.append(partial);
+ buffer_.append(block);
+ // XXX Hopefully this upholds for all std::string implementations
+ DCHECK_GE(buffer_.capacity() - buffer_.size(), simdjson::SIMDJSON_PADDING);
Review Comment:
`simdjson::padded_input` will provide a (possibly) copy-free version, but it
is not yet in a released version:
https://github.com/simdjson/simdjson/commit/85cadf4074b78c7dfa7dabe89bef373b84cf0ba6
Also note that their code is more involved than the checking you do here:
```
const size_t len = s.size();
const size_t cap = s.capacity();
// Here we have the string content from data() to data() + size(),
// but the memory is accessible from data() to data() + capacity().
const size_t needed_padding = (cap - len) < simdjson::SIMDJSON_PADDING
? simdjson::SIMDJSON_PADDING - (cap - len) : 0;
if (needed_padding > 0 && needs_allocation(s.data(), cap, needed_padding))
{
storage = simdjson::padded_string(s);
} else {
storage = simdjson::padded_string_view(
s.data(), len, len + simdjson::SIMDJSON_PADDING);
}
```
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
