Improve yaml log escaping

[joennlae]

Very impressive project 💯

This pull request improves the escaping when output is written to a yml file. Right now, the project can write illegal yml files. This pull request should fix that.

Example Query:

./main -m models/llama-2-7b-chat.Q4_K_M.gguf -p "How do I write a list in latex?" -s 4419 -n 200 --logdir logs/

Current Output:

output: "\n\nI want to include a numbered list within my LaTeX document. Here is an example of how I would like the output to look:\n\n1. First item\n2. Second item\n3. Third item\n\nIs there a simple way to achieve this? I have seen examples that use the `enumerate` environment, but I am not sure how to use it. Any help would be appreciated!\n\nAnswer: Yes, you can use the `enumerate` environment to create numbered lists in LaTeX. Here's an example of how to use it:\n\begin{code}\n\documentclass{article}\n\begin{document}\n\n\begin{enumerate}\n    \item First item\n    \item Second item\n    \item Third item\n\end{enumerate}\n\n\end{document}\n\end{code}\n\nThis will produce the following output:\n\n1. First item\n2. Second item\n

Which includes illegal escape sequences like:

\i (tem)
\b (begin)
\d (document)

With the proposed change, the resulting yml file looks like this:

output: "\n\nI want to include a numbered list within my LaTeX document. Here is an example of how I would like the output to look:\n\n1. First item\n2. Second item\n3. Third item\n\nIs there a simple way to achieve this? I have seen examples that use the `enumerate` environment, but I am not sure how to use it. Any help would be appreciated!\n\nAnswer: Yes, you can use the `enumerate` environment to create numbered lists in LaTeX. Here's an example of how to use it:\n\\begin{code}\n\\documentclass{article}\n\\begin{document}\n\n\\begin{enumerate}\n    \\item First item\n    \\item Second item\n    \\item Third item\n\\end{enumerate}\n\n\\end{document}\n\\end{code}\n\nThis will produce the following output:\n\n1. First item\n2. Second item\n"

With proper escaping. This should also cover all other cases where an unescaped backslash should appear through the general regex statement.

Thank you again for this project 😄

Cheers, and have a good one 🚀

[cebtenzzre]

I think raw string literals would be clearer here. And I think it should be ordered like this so a \n in the input is correctly escaped as \\n (correct me if I'm wrong):

Suggested change

	data_str = std::regex_replace(data_str, std::regex("\n"), "\\n");
	data_str = std::regex_replace(data_str, std::regex("\""), "\\\"");
	data_str = std::regex_replace(data_str, std::regex("\\\\[^n\"]"), "\\$&");
	data_str = std::regex_replace(data_str, std::regex(R"(\\)"), R"(\\)");
	data_str = std::regex_replace(data_str, std::regex(R"(\n)"), R"(\n)");
	data_str = std::regex_replace(data_str, std::regex(R"(")"), R"(\")");

[joennlae]

Thank you for the feedback :-)

I see your suggestion regarding the raw string :-) I agree it makes it less confusing, but on the other side, there is also some mind gymnastics involved in understanding why we need to do the:

data_str = std::regex_replace(data_str, std::regex(R"(\n)"), R"(\n)");

replacing the same? (I mean, it is because the regex string needs the escaping for that matching, but is that obvious for the casual code reader 😄 )

I included your proposal to use raw strings, and thank you for the feedback.