libcruft-util/uri.cpp

#include "./uri.hpp"

#include "./cast.hpp"
#include "./string.hpp"

#include "./debug/panic.hpp"
#include "./debug/assert.hpp"
#include "./debug/warn.hpp"

#include <ostream>

using cruft::uri;


///////////////////////////////////////////////////////////////////////////////
static void
minimal_percent_encode (std::string &str)
{
    for (ssize_t i = 0; i < std::ssize (str); ++i) {
        if (str[i] == ' ')
            str.replace (i, 1, "%20");
    }
}


///////////////////////////////////////////////////////////////////////////////
cruft::uri::uri (std::string &&_value):
    m_offsets {},
    m_value (std::move (_value))
{
    minimal_percent_encode (m_value);

    parse ();

    // Ensure the offsets are well-ordered.
    //
    // We want the QUERY offsets to come after the PATH offsets even if there
    // is no QUERY data (or at least be placed coincident with the PATH
    // offsets).
    for (int i = 1; i < NUM_COMPONENTS; ++i) {
        // We very specifically do not want to move any offset that has a
        // non-zero (ie, found) offset.
        //
        // If we do, this prevents us from computing `pqf` and retaining a
        // trailing "?" for an empty query.
        if (m_offsets[i].second or m_offsets[i].first)
            continue;

        m_offsets[i] = {
            m_offsets[i - 1].second,
            m_offsets[i - 1].second
        };
    }

    CHECK_SANITY (*this);
}


//-----------------------------------------------------------------------------
cruft::uri::uri (const char *str):
    uri (std::string (str))
{ ; }


//-----------------------------------------------------------------------------
uri::uri (cruft::view<const char *> _value):
    uri (std::string (_value.begin (), _value.end ()))
{ ; }


//-----------------------------------------------------------------------------
uri::uri (const std::string &_value):
    uri (std::string (_value))
{ ; }


//-----------------------------------------------------------------------------
static std::string
combine_components (
    std::string_view scheme,
    std::string_view authority,
    std::string_view path,
    std::string_view query,
    std::string_view fragment
) {
    std::string res;
    res.reserve (
        scheme.size () +
        strlen ("://") + authority.size () +
        path.size () +
        strlen ("?") + query.size () +
        strlen ("#") + fragment.size ()
    );

    if (!scheme.empty ()) {
        res += scheme;
        res += ":";
    }

    if (!authority.empty ()) {
        res += "//";
        res += authority;
    }

    res += path;

    if (!query.empty ()) {
        res += "?";
        res += query;
    }

    if (!fragment.empty ()) {
        res += "#";
        res += fragment;
    }

    return res;
}


//-----------------------------------------------------------------------------
uri::uri (
    std::string_view scheme,
    std::string_view authority,
    std::string_view path,
    std::string_view query,
    std::string_view fragment
) : uri (combine_components (scheme, authority, path, query, fragment))
{ ; }


///////////////////////////////////////////////////////////////////////////////
std::string_view
uri::heirarchical (void) const&
{
    for (int i = USER; i <= PATH; ++i)
        if (m_offsets[i].first != m_offsets[i].second)
            return {
                m_value.data () + m_offsets[   i].first,
                m_value.data () + m_offsets[PATH].second
            };
    return { m_value.data () + m_offsets[USER].first, 0 };
}


//-----------------------------------------------------------------------------
std::string_view
uri::authority (void) const&
{
    for (int i = USER; i <= PORT; ++i)
        if (m_offsets[i].first != m_offsets[i].second)
            return {
                m_value.data () + m_offsets[   i].first,
                m_value.data () + m_offsets[PORT].second
            };

    return {
        m_value.data () + m_offsets[USER].first, 0
    };
}


///////////////////////////////////////////////////////////////////////////////
std::string_view
uri::get (component c) const&
{
    CHECK_INDEX (c, NUM_COMPONENTS);
    return {
        m_value.data () + m_offsets[c].first,
        m_value.data () + m_offsets[c].second,
    };
}


//-----------------------------------------------------------------------------
void
uri::set (component c, std::string_view val)
{
    std::ptrdiff_t const diff = val.size () - (m_offsets[c].second - m_offsets[c].first);

    m_value.replace (
        m_offsets[c].first,
        m_offsets[c].second - m_offsets[c].first,
        val
    );

    m_offsets[c].second = cruft::cast::lossless<int> (m_offsets[c].first + val.size ());

    for (int i = c + 1; i != component::NUM_COMPONENTS; ++i) {
        m_offsets[i].first  += diff;
        m_offsets[i].second += diff;
    }

    CHECK_SANITY (*this);
}


//-----------------------------------------------------------------------------
void uri::clear_fragment ()
{
    if (m_offsets[FRAGMENT].first == m_offsets[FRAGMENT].second)
        return;

    m_value.erase (
        m_offsets[FRAGMENT].first,
        m_offsets[FRAGMENT].second - m_offsets[FRAGMENT].first
    );
    CHECK (m_value.back () == '#');
    m_value.pop_back ();

    // Don't set this to nullptr. Other code assumes these views aren't null
    // (eg, for offsetting during copy construction).
    m_offsets[FRAGMENT].first = m_offsets[FRAGMENT].second = m_offsets[FRAGMENT - 1].second;
}


///////////////////////////////////////////////////////////////////////////////
std::vector<std::pair<std::string, std::string>>
cruft::query_to_vector (std::string_view val)
{
    std::vector<std::pair<std::string, std::string>> res;

    for (auto const tok: cruft::tokeniser (val, '&')) {
        auto const &[k, v] = cruft::split_on (tok, '=');
        res.emplace_back (
            std::string (k.begin (), k.size ()),
            std::string (v.begin (), v.size ())
        );
    }

    return res;
}


//-----------------------------------------------------------------------------
std::string
cruft::vector_to_query (std::vector<std::pair<std::string, std::string>> const &val)
{
    // Test for empty up front so that we can simplify the string
    // concatenation below.
    if (val.empty ())
        return "";

    std::string res;
    for (auto const &[k, v]: val) {
        res += k;
        res += '=';
        res += v;
        res += '&';
    }

    // The string must be non-zero length because we've tested for the empty
    // set initially.
    CHECK (!res.empty ());
    res.resize (res.size () - 1);
    return res;
}


///////////////////////////////////////////////////////////////////////////////
std::array<std::string_view, cruft::uri::NUM_COMPONENTS>
uri::components () const& noexcept
{
    std::array<std::string_view, cruft::uri::NUM_COMPONENTS> res;
    for (int i = 0; i != NUM_COMPONENTS; ++i)
        res[i] = get (uri::component (i));
    return res;
}



///////////////////////////////////////////////////////////////////////////////
bool
cruft::operator== (cruft::uri const &a, cruft::uri const &b) noexcept
{
    return a.value () == b.value ();
}


///////////////////////////////////////////////////////////////////////////////
static uint8_t
hex_to_uint (char c)
{
    if (c >= '0' && c <= '9')
        return c - '0';

    if (c >= 'A' && c <= 'F')
        return c - 'A' + 10;

    if (c >= 'a' && c <= 'f')
        return c - 'a' + 10;

    unreachable ();
}


//-----------------------------------------------------------------------------
std::string
cruft::uri::percent_decode (view<const char*> s)
{
    if (s.size () == 0)
        return std::string ();

    // Early check for late percent-encoding so we can simplify the decode loop
    {
        auto tail = std::find (s.size () < 3 ? s.begin ()
                                             : s.end () - 2,
                               s.end (),
                               '%');
        if (tail != s.end ())
            throw parse_error ("triple overlaps end");
    }

    // Allocate and size a potentially overlong output string. This allows us
    // to copy directly into its buffer. We'll shorten it at the end.
    std::string out;
    out.resize (s.size ());

    // Find the percent, copy until that, decode, advance, repeat.
    auto out_cursor = out.begin ();

    for (auto i = s.begin (); i < s.end (); ++i) {
        auto cursor = std::find (i, s.end (), '%');

        if (cursor == s.end ()) {
            out_cursor = std::copy (i, s.end (), out_cursor);
            break;
        }

        out_cursor = std::copy (i, cursor, out_cursor);
        *out_cursor = hex_to_uint (cursor[1]) << 4 | hex_to_uint(cursor[2]);

        i += 3;
    }

    out.resize (out.end () - out_cursor);
    return out;
}



///////////////////////////////////////////////////////////////////////////////
std::ostream&
cruft::operator<< (std::ostream &os, cruft::uri::component c)
{
    switch (c) {
        case cruft::uri::SCHEME:         return os << "SCHEME";
        case cruft::uri::USER:           return os << "USER";
        case cruft::uri::HOST:           return os << "HOST";
        case cruft::uri::PORT:           return os << "PORT";
        case cruft::uri::PATH:           return os << "PATH";
        case cruft::uri::QUERY:          return os << "QUERY";
        case cruft::uri::FRAGMENT:       return os << "FRAGMENT";

        case cruft::uri::NUM_COMPONENTS:
            unreachable ();
    }

    unreachable ();
}


//-----------------------------------------------------------------------------
std::ostream&
cruft::operator<< (std::ostream &os, cruft::uri const &val)
{
    return os << val.value ();
}


///////////////////////////////////////////////////////////////////////////////
static std::string
merge (std::string_view base, std::string_view relative)
{
    auto const slash = std::find (std::rbegin (base), std::rend (base), '/');
    if (slash == std::rend (base))
        return fmt::format ("/{}", relative);

    return fmt::format (
        "{}/{}",
        std::string_view (base.begin (), std::distance (base.begin (), slash.base ()) - 1),
        relative
    );
}


//-----------------------------------------------------------------------------
static std::string
remove_dot_segments (std::string_view path)
{
    std::vector<std::string_view> src;
    for (auto const &i: cruft::tokeniser (path, '/'))
        src.push_back (std::string_view (i.begin (), i.size ()));

    bool const absolute = !path.empty () && path[0] == '/';
    bool const trailing = !src.empty () && (src.back () == "" or src.back () == "." or src.back () == "..");

    std::vector<std::string_view> dst;
    for (auto const &i: src) {
        if (i == "..") {
            if (!dst.empty ()) {
                if (dst.back () == "..")
                    dst.push_back (i);
                else
                    dst.pop_back ();
            }
        } else if (i != "." and i != "") {
            dst.push_back (i);
        }
    }

    std::string res = absolute ? "/" : "";
    for (auto const &i: dst) {
        res.append (i);
        res.append ("/");
    }

    if (!trailing)
        if (!res.empty ())
            res.pop_back ();

    return res;
}


//-----------------------------------------------------------------------------
// Uniform Resource Identifier (URI): Generic Syntax
// https://www.ietf.org/rfc/rfc3986.txt
// 5.2.  Relative Resolution
cruft::uri
cruft::resolve (cruft::uri const &base, cruft::uri const &relative)
{
    using namespace std::string_literals;

    if (!relative.scheme ().empty ())
        return relative;

    std::string_view scheme    = base.scheme ();
    std::string_view authority = base.authority ();
    std::string      path      = std::string (base.path ());
    std::string_view query     = base.query ();
    std::string_view fragment  = base.fragment ();

    if (!relative.scheme ().empty ()) {
        scheme    = relative.scheme ();
        authority = relative.authority ();
        path      = remove_dot_segments (relative.path ());
        query     = relative.query ();
    } else {
        if (!relative.authority ().empty ()) {
            authority = relative.authority ();
            path      = remove_dot_segments(relative.path ());
            query     = relative.query ();
        } else {
            if (relative.path ().empty ()) {
                path = base.path ();
                if (!relative.query ().empty ())
                    query = relative.query ();
                else
                    query = base.query ();
            } else {
                if (relative.path ().starts_with ("/"))
                    path = remove_dot_segments(relative.path ());
                else {
                    if (!base.authority ().empty () and base.path ().empty ())
                        path = fmt::format ("/{}", relative.path ());
                    else
                        path = merge (base.path (), relative.path ());
                    path = remove_dot_segments (path);
                }
                query = relative.query ();
            }
            authority = base.authority ();
        }
        scheme = base.scheme ();
    }

    fragment = relative.fragment ();

    return { scheme, authority, path, query, fragment };
}


///////////////////////////////////////////////////////////////////////////////
cruft::uri
cruft::normalise (cruft::uri const &src)
{
    auto res = src;
    res.set (
        uri::component::PATH,
        remove_dot_segments (src.path ())
    );
    return res;
}


///////////////////////////////////////////////////////////////////////////////
template <>
bool
cruft::debug::validator<cruft::uri>::is_valid (cruft::uri const &val) noexcept
{
    auto const &value = val.value ();
    auto const &components = val.components ();

    // Each component should fall within the value string
    for (auto const &i: components) {
        RETURN_FALSE_UNLESS (i.begin () >= value.data ());
        RETURN_FALSE_UNLESS (i.end   () <= value.data () + value.size ());
    }

    for (int i = 1; i < uri::NUM_COMPONENTS; ++i) {
        auto const &a = components[i - 1];
        auto const &b = components[i    ];

        // All of our data needs to occur before the next component
        //
        // It may not be coincident with a sibling if there is padding between
        // (as is the case directly after the scheme).
        RETURN_FALSE_UNLESS (a.begin () <= b.begin ());
        RETURN_FALSE_UNLESS (a.end   () <= b.end   ());
        RETURN_FALSE_UNLESS (a.end   () <= b.begin ());
    }

    return true;
}