What is the best way to read files from ifstream in terms of performance.

[zhoupengwei]

Hi author, I am using rapidyaml to read and write content from yaml files. For reading the content from file based on the ifstream, as follows:

// open the ifstream
std::ifstream file;
// convert to string
std::string file_content((std::istreambuf_iterator<char>(file)),
                           std::istreambuf_iterator<char>());
// convert to ryml
ryml::Tree tree = ryml::parse_in_arena(ryml::to_csubstr(file_content));

This method involves additional copies. For writing the content into file based on the ostream, as follows:

std::ofstream file;
ryml::Tree tree;
file << tree;

This way uses the operator overloaded by ryml, It is very convenient!

But, ryml can not use the operator overloaded to read file content from ifstream, like the output

std::ifstream file;
// error
file >> tree;

So how to construct a tree from ifstream to ensure the best performance and avoid copying.

[biojppm]

Well, for best performance you should not use iostreams, period. They have generally terrible performance.

If you must use them, however, then your first method is very close: create a string from the stream and then read the string. But you can use an existing tree! Do this instead:

// file_content as above

// no copies:
ryml::Tree tree;

// minimize resizes during parsing:
tree.reserve(suitable_number_of_nodes);
tree.reserve_arena(suitable_arena_size);

// now parse into the existing tree, and save one copy by parsing in place"
ryml::parse_in_place(ryml::to_substr(file_content), &tree);

[zhoupengwei]

@biojppm Thank you for your reply, what is the best way to save and read rapidyaml into regular yaml file?

[biojppm]

For parsing, ryml reads only from existing memory buffers. So on this subject your question is how to read a file into memory. That question is outside the scope of ryml. Maybe have a look at https://stackoverflow.com/questions/2912520/read-file-contents-into-a-string-in-c

As for emitting, ryml lets you emit YAML to memory, file or iostream. Refer to the emit documentation to find out the appropriate call.

[zhoupengwei]

@biojppm Thank you for the detail response, my problem have been solved.