json_view: exceptions name the path and the bytes of the value

With JSON_DIAGNOSTICS and JSON_DIAGNOSTIC_POSITIONS, the exceptions of views
now carry the same context as those of basic_json: "(/path) " and
"(bytes a-b) " in front of the message, in basic_json's format (with
JSON_DIAGNOSTICS, no context for the root). Before, views threw without any
context, and since get<T>() reads the view directly, conversions had lost
even the relative path they used to get through materialize().

The throw helpers of detail/view/errors.hpp take an error_site (document
and node; empty without the two macros). Only on the error path, the path is
found by descending from the root (pruned by subtree extent before the first
edit, a full walk through edited sequences after), and the bytes as
materialize() sets them (position_source; a container's end by following the
chain of last elements and scanning to the closing brackets). from_json() for
views and the strict enum macro put the same context in front through
exception_prefix(). Iterators keep their container only under the two
macros, for key(). Without the macros, the accessors compile to the same
code as before.

Tests compare every value's exceptions with those of the parsed or
materialized basic_json value, including edited documents; the checks that
were disabled under the two macros run again.

Signed-off-by: Niels Lohmann <mail@nlohmann.me>

include/nlohmann/detail/view/errors.hpp

@@ -10,11 +10,15 @@
 
 #include <algorithm> // min
 #include <cstddef> // size_t
-#include <string> // string
+#include <string> // string, to_string
+#include <vector> // vector
 
 #include <nlohmann/json.hpp>
 #include <nlohmann/detail/view/builder.hpp>
+#include <nlohmann/detail/view/document_data.hpp>
 #include <nlohmann/detail/view/macro_scope.hpp>
+#include <nlohmann/detail/view/materialize.hpp>
+#include <nlohmann/detail/view/node.hpp>
 
 NLOHMANN_JSON_NAMESPACE_BEGIN
 namespace detail
@@ -22,8 +26,277 @@ namespace detail
 namespace view
 {
 
+/*!
+@brief the value an exception is about
+
+basic_json passes the value to the exception, which then names its path
+(JSON_DIAGNOSTICS) and its bytes in the source (JSON_DIAGNOSTIC_POSITIONS). A
+view's value has no parent pointer and no stored positions: both are found
+from the document and the node, when the exception is thrown (so only the
+error path pays for them). Without the two macros, a site is empty.
+*/
+class error_site
+{
+  public:
+    error_site() noexcept = default;
+
+    error_site(const document_data* d, const node* n) noexcept
+#if JSON_DIAGNOSTICS || JSON_DIAGNOSTIC_POSITIONS
+        : m_doc(d), m_node(n)
+#endif
+    {
+        static_cast<void>(d);
+        static_cast<void>(n);
+    }
+
+#if JSON_DIAGNOSTICS || JSON_DIAGNOSTIC_POSITIONS
+    const document_data* doc() const noexcept
+    {
+        return m_doc;
+    }
+    const node* value() const noexcept
+    {
+        return m_node;
+    }
+
+  private:
+    const document_data* m_doc = nullptr;
+    const node* m_node = nullptr;
+#endif
+};
+
+/// the site of a view's value (the views make this a friend)
+struct view_access
+{
+    template<typename View>
+    static error_site site(const View& v) noexcept
+    {
+        return error_site(v.m_doc, v.m_node);
+    }
+};
+
+template<typename View>
+error_site site_of(const View& v) noexcept
+{
+    return view_access::site(v);
+}
+
+#if JSON_DIAGNOSTICS
+/*!
+@brief the JSON pointer tokens from the root of document @a d to node @a target
+
+Nodes have no parent index, so the path is found by descending from the root:
+before the first edit, only into the subtree that contains the target (a
+container spans the next `next` nodes); after edits, elements can live in
+other sequences, so every container is searched. Iterative (the nesting depth
+is limited by memory only). False if the target is not found (then tokens is
+unspecified).
+*/
+template<bool Editable>
+bool value_path(const document_data& d, const node* target, std::vector<std::string>& tokens)
+{
+    using nav = navigation<Editable>;
+    struct frame
+    {
+        const node* pos;
+        const node* end;
+        std::size_t index;
+        bool object;
+    };
+    const node* const root = d.nodes;
+    if (target == root)
+    {
+        return true;
+    }
+    if (!is_container(*root))
+    {
+        return false;
+    }
+    std::vector<frame> stack;
+    stack.push_back(frame{nav::first(d, root), nav::end(d, root), 0, root->kind == node_kind::object});
+    // (tokens[i] leads from stack[i] to stack[i + 1], or to the target)
+    while (!stack.empty())
+    {
+        frame& f = stack.back();
+        if (f.pos == f.end)
+        {
+            stack.pop_back();
+            if (!tokens.empty())
+            {
+                tokens.pop_back();
+            }
+            continue;
+        }
+        const node* entry = f.pos;
+        std::string token;
+        if (f.object)
+        {
+            // (as basic_json: the key up to a null byte)
+            token.assign(d.str(*entry), entry->len);
+            const std::size_t nul = token.find('\0');
+            if (nul != std::string::npos)
+            {
+                token.resize(nul);
+            }
+            ++entry;
+        }
+        else
+        {
+            token = std::to_string(f.index);
+        }
+        const node* const child = nav::value(entry);
+        f.pos = document_data::after(entry);
+        ++f.index;
+        if (child == target)
+        {
+            tokens.push_back(token);
+            return true;
+        }
+        if (is_container(*child) && (nav::editable() || (target > child && target < child + child->next)))
+        {
+            tokens.push_back(token);
+            stack.push_back(frame{nav::first(d, child), nav::end(d, child), 0, child->kind == node_kind::object});
+        }
+    }
+    return false;
+}
+#endif
+
+#if JSON_DIAGNOSTIC_POSITIONS
+/*!
+@brief the bytes [start, end) of node @a n in the source, as materialize() sets them
+
+Scalars as position_source::set_scalar; an array or object from its opening
+bracket to its closing bracket, which is found by following the chain of last
+elements to a scalar (or an empty container) and scanning from there to the
+closing brackets. None (npos) for values written by edits, and for containers
+whose last element has none, as with materialize().
+*/
+template<bool Editable>
+void value_range(const document_data& d, const node* n, std::size_t& start, std::size_t& end)
+{
+    using nav = navigation<Editable>;
+    position_source p;
+    start = std::string::npos;
+    end = std::string::npos;
+    if (!is_container(*n))
+    {
+        p.set_scalar(d, *n);
+        start = p.start;
+        end = p.end;
+        return;
+    }
+    if ((n->flags & node_flags::is_new) != 0 || n->off >= d.size)
+    {
+        return;
+    }
+    // the open brackets: n, then the last element as long as it is a non-empty container
+    std::size_t closing = 0;
+    std::size_t pos = std::string::npos;
+    for (const node* c = n;;)
+    {
+        if (is_container(*c))
+        {
+            if ((c->flags & node_flags::is_new) != 0 || c->off >= d.size)
+            {
+                return;
+            }
+            ++closing;
+            const node* const first = nav::first(d, c);
+            const node* const last_end = nav::end(d, c);
+            if (first == last_end)
+            {
+                pos = std::size_t{c->off} + 1;
+                break;
+            }
+            const std::size_t value_offset = c->kind == node_kind::object ? 1 : 0;
+            const node* last = first;
+            for (const node* e = document_data::after(first + value_offset); e != last_end; e = document_data::after(e + value_offset))
+            {
+                last = e;
+            }
+            c = nav::value(last + value_offset);
+            continue;
+        }
+        p.set_scalar(d, *c);
+        pos = p.end;
+        break;
+    }
+    for (; closing != 0 && pos != std::string::npos; --closing)
+    {
+        pos = position_source::container_end(d.base[0], d.size, pos);
+    }
+    if (pos != std::string::npos)
+    {
+        start = n->off;
+        end = pos;
+    }
+}
+#endif
+
+/// the context basic_json puts in front of the message of an exception about
+/// a value: "(/path) " with JSON_DIAGNOSTICS (none for the root), then
+/// "(bytes a-b) " with JSON_DIAGNOSTIC_POSITIONS (with JSON_DIAGNOSTICS, only
+/// if there is a path)
+inline std::string diagnostics(const error_site site)
+{
+#if JSON_DIAGNOSTICS || JSON_DIAGNOSTIC_POSITIONS
+    if (site.doc() == nullptr || site.value() == nullptr)
+    {
+        return "";
+    }
+    const document_data& d = *site.doc();
+    const bool edited = d.edits != nullptr;
+    std::string bytes;
+#if JSON_DIAGNOSTIC_POSITIONS
+    std::size_t start = 0;
+    std::size_t end = 0;
+    if (edited)
+    {
+        value_range<true>(d, site.value(), start, end);
+    }
+    else
+    {
+        value_range<false>(d, site.value(), start, end);
+    }
+    if (start != std::string::npos && end != std::string::npos)
+    {
+        bytes = concat("(bytes ", std::to_string(start), "-", std::to_string(end), ") ");
+    }
+#endif
+#if JSON_DIAGNOSTICS
+    std::vector<std::string> tokens;
+    const bool found = edited ? value_path<true>(d, site.value(), tokens) : value_path<false>(d, site.value(), tokens);
+    if (!found || tokens.empty())
+    {
+        return "";
+    }
+    std::string path;
+    for (const std::string& token : tokens)
+    {
+        path.push_back('/');
+        path += detail::escape(token);
+    }
+    return concat('(', path, ") ", bytes);
+#else
+    return bytes;
+#endif
+#else
+    static_cast<void>(site);
+    return "";
+#endif
+}
+
+/// the context of an exception about view @a v (see exception_prefix())
+template<typename View>
+std::string diagnostics_of(const View& v)
+{
+    return diagnostics(site_of(v));
+}
+
 // Exceptions are thrown out of line, so that the accessors that may throw stay
-// small enough to be inlined.
+// small enough to be inlined. The overloads with an error_site put its context
+// in front of the message.
 
 [[noreturn]] NLOHMANN_VIEW_NOINLINE inline void throw_type_error(exception_id id, const char* prefix, const char* type)
 {
@@ -45,6 +318,21 @@ namespace view
     NLOHMANN_VIEW_THROW(invalid_iterator::create(id, msg, nullptr));
 }
 
+[[noreturn]] NLOHMANN_VIEW_NOINLINE inline void throw_type_error(const error_site site, exception_id id, const char* prefix, const char* type)
+{
+    NLOHMANN_VIEW_THROW(type_error::create(id, concat(diagnostics(site), prefix, type), nullptr));
+}
+
+[[noreturn]] NLOHMANN_VIEW_NOINLINE inline void throw_out_of_range(const error_site site, exception_id id, const std::string& msg)
+{
+    NLOHMANN_VIEW_THROW(out_of_range::create(id, concat(diagnostics(site), msg), nullptr));
+}
+
+[[noreturn]] NLOHMANN_VIEW_NOINLINE inline void throw_invalid_iterator(const error_site site, exception_id id, const char* msg)
+{
+    NLOHMANN_VIEW_THROW(invalid_iterator::create(id, concat(diagnostics(site), msg), nullptr));
+}
+
 /// a parse error without a position (as those of json_pointer)
 [[noreturn]] NLOHMANN_VIEW_NOINLINE inline void throw_parse_error(exception_id id, const std::string& msg)
 {
@@ -53,22 +341,22 @@ namespace view
 
 /// out_of_range.401: the array index @a idx is not below the size
 template<typename Index>
-[[noreturn]] NLOHMANN_VIEW_NOINLINE void throw_array_index_out_of_range(const Index idx)
+[[noreturn]] NLOHMANN_VIEW_NOINLINE void throw_array_index_out_of_range(const error_site site, const Index idx)
 {
-    throw_out_of_range(exception_id::array_index_out_of_range, concat("array index ", std::to_string(idx), " is out of range"));
+    throw_out_of_range(site, exception_id::array_index_out_of_range, concat("array index ", std::to_string(idx), " is out of range"));
 }
 
 /// out_of_range.402: the JSON Pointer token '-' of an array of @a size elements
-[[noreturn]] NLOHMANN_VIEW_NOINLINE inline void throw_array_index_past_end(std::size_t size)
+[[noreturn]] NLOHMANN_VIEW_NOINLINE inline void throw_array_index_past_end(const error_site site, std::size_t size)
 {
-    throw_out_of_range(exception_id::pointer_past_the_end_index, concat("array index '-' (", std::to_string(size), ") is out of range"));
+    throw_out_of_range(site, exception_id::pointer_past_the_end_index, concat("array index '-' (", std::to_string(size), ") is out of range"));
 }
 
 /// out_of_range.403: the object has no member @a key
 template<typename Key>
-[[noreturn]] NLOHMANN_VIEW_NOINLINE void throw_key_not_found(const Key& key)
+[[noreturn]] NLOHMANN_VIEW_NOINLINE void throw_key_not_found(const error_site site, const Key& key)
 {
-    throw_out_of_range(exception_id::key_not_found, concat("key '", key, "' not found"));
+    throw_out_of_range(site, exception_id::key_not_found, concat("key '", key, "' not found"));
 }
 
 /// SAX handler of the error replay: accepts every event without storing it,