Correct way to convert http::response_parser<http::buffer_body> to http::response_parser<http::string_body>

Question

I have a response parser of type http::response_parser<http::buffer_body>, which I need to convert to a parser of type http::response_parser<http::string_body>. I do the conversion after reading the headers via http::async_read_header.

At the time of conversion, http::response_parser<http::buffer_body> does not have a pointer to the buffer, since the buffer is not required to read the headers.

In code it looks like this

auto parser = std::make_shared<boost::beast::http::response_parser<boost::beast::http::buffer_body>>();

boost::beast::http::async_read_header(m_stream, m_buffer, *parser,
    [this, self, parser](const boost::system::error_code& ec, std::size_t bytes_transferred)
    {
        auto new_parser = std::make_shared<boost::beast::http::response_parser<boost::beast::http::string_body>>(std::move(*parser));
    });

My question is, is it safe to do such conversion while http::response_parser<http::buffer_body> doesn't have a buffer pointer?

The concern is that when creating a new parser, I think the converting constructor might try to access a buffer_body buffer that has not been set.

Should I set a valid buffer_body pointer before converting, or at least set the buffer size to 0?

sehe · Accepted Answer · 2025-07-05 14:39:38Z

2

In my comment here:

Also, use this constructor to switch the body type on the fly (after parsing headers). See also https://live.boost.org/doc/libs/1_88_0/libs/beast/doc/html/beast/more_examples/change_body_type.html

I linked both the constructor overload and the documentation article which contains a concrete example of when and how to use it.

1. My question is, is it safe to do such conversion while `http::response_parser<http::buffer_body>` doesn't have a buffer pointer?

The parser does not "have" a buffer to begin with.

2. The concern is that when creating a new parser, I think the converting constructor might try to access a buffer_body buffer that has not been set.

The constructor overload specifically, and by definition, does NOT assume a body:

This constructs a new parser by move constructing the header from another parser with a different body type. The constructed-from parser must not have any parsed body octets or initialized BodyReader, otherwise an exception is generated.

That's in the documentation of that overload.

answered Jul 5 at 14:39

sehe

400k49 gold badges475 silver badges673 bronze badges

Sign up to request clarification or add additional context in comments.

5 Comments

Joe J Jul 5 at 14:54

Thanks! I read this documentation. I am confused by this quote: "The constructed-from parser must not have any parsed body octets ... otherwise an exception is generated". For this reason, I think that the constructor can access the parser body to check if it has read bytes in order to find out whether to throw an exception or not. So, the conversion I specified can be done safely?

sehe Jul 5 at 15:47

On the contrary. The constructor makes NO sense if the body reader was initialized. Which is why the pre-condition checks it. Yes, it's safe, otherwise the documentation example wouldn't do exactly the same thing...

sehe Jul 9 at 16:52

@JoeJ is this one yours? github.com/boostorg/beast/issues/3017

Joe J Jul 14 at 9:53

No (...........)

sehe Jul 14 at 10:16

Well, regardless, the devs have confirmed this answer by now

Collectives™ on Stack Overflow

Correct way to convert http::response_parser<http::buffer_body> to http::response_parser<http::string_body>

1 Answer 1

1. My question is, is it safe to do such conversion while `http::response_parser<http::buffer_body>` doesn't have a buffer pointer?

2. The concern is that when creating a new parser, I think the converting constructor might try to access a buffer_body buffer that has not been set.

5 Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

1. My question is, is it safe to do such conversion while http::response_parser<http::buffer_body> doesn't have a buffer pointer?

2. The concern is that when creating a new parser, I think the converting constructor might try to access a buffer_body buffer that has not been set.

5 Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related

1. My question is, is it safe to do such conversion while `http::response_parser<http::buffer_body>` doesn't have a buffer pointer?