use byte indexed locations for PosIdx

we now keep not a table of all positions, but a table of all origins and
their sizes. position indices are now direct pointers into the virtual
concatenation of all parsed contents. this slightly reduces memory usage
and time spent in the parser, at the cost of not being able to report
positions if the total input size exceeds 4GiB. this limit is not unique
to nix though, rustc and clang also limit their input to 4GiB (although
at least clang refuses to process inputs that are larger, we will not).

this new 4GiB limit probably will not cause any problems for quite a
while, all of nixpkgs together is less than 100MiB in size and already
needs over 700MiB of memory and multiple seconds just to parse. 4GiB
worth of input will easily take multiple minutes and over 30GiB of
memory without even evaluating anything. if problems *do* arise we can
probably recover the old table-based system by adding some tracking to
Pos::Origin (or increasing the size of PosIdx outright), but for time
being this looks like more complexity than it's worth.

since we now need to read the entire input again to determine the
line/column of a position we'll make unsafeGetAttrPos slightly lazy:
mostly the set it returns is only used to determine the file of origin
of an attribute, not its exact location. the thunks do not add
measurable runtime overhead.

notably this change is necessary to allow changing the parser since
apparently nothing supports nix's very idiosyncratic line ending choice
of "anything goes", making it very hard to calculate line/column
positions in the parser (while byte offsets are very easy).

src/libexpr/pos-table.hh

@@ -7,6 +7,7 @@
 #include "chunked-vector.hh"
 #include "pos-idx.hh"
 #include "position.hh"
+#include "sync.hh"
 
 namespace nix {
 
@@ -17,66 +18,69 @@ public:
     {
         friend PosTable;
     private:
-        // must always be invalid by default, add() replaces this with the actual value.
-        // subsequent add() calls use this index as a token to quickly check whether the
-        // current origins.back() can be reused or not.
-        mutable uint32_t idx = std::numeric_limits<uint32_t>::max();
+        uint32_t offset;
 
-        // Used for searching in PosTable::[].
-        explicit Origin(uint32_t idx)
-            : idx(idx)
-            , origin{std::monostate()}
-        {
-        }
+        Origin(Pos::Origin origin, uint32_t offset, size_t size):
+            offset(offset), origin(origin), size(size)
+        {}
 
     public:
         const Pos::Origin origin;
+        const size_t size;
 
-        Origin(Pos::Origin origin)
-            : origin(origin)
+        uint32_t offsetOf(PosIdx p) const
         {
+            return p.id - 1 - offset;
         }
     };
 
-    struct Offset
-    {
-        uint32_t line, column;
-    };
-
 private:
-    std::vector<Origin> origins;
-    ChunkedVector<Offset, 8192> offsets;
+    using Lines = std::vector<uint32_t>;
 
-public:
-    PosTable()
-        : offsets(1024)
-    {
-        origins.reserve(1024);
-    }
+    std::map<uint32_t, Origin> origins;
+    mutable Sync<std::map<uint32_t, Lines>> lines;
 
-    PosIdx add(const Origin & origin, uint32_t line, uint32_t column)
+    const Origin * resolve(PosIdx p) const
     {
-        const auto idx = offsets.add({line, column}).second;
-        if (origins.empty() || origins.back().idx != origin.idx) {
-            origin.idx = idx;
-            origins.push_back(origin);
-        }
-        return PosIdx(idx + 1);
-    }
+        if (p.id == 0)
+            return nullptr;
 
-    Pos operator[](PosIdx p) const
-    {
-        if (p.id == 0 || p.id > offsets.size())
-            return {};
         const auto idx = p.id - 1;
         /* we want the last key <= idx, so we'll take prev(first key > idx).
-           this is guaranteed to never rewind origin.begin because the first
-           key is always 0. */
-        const auto pastOrigin = std::upper_bound(
-            origins.begin(), origins.end(), Origin(idx), [](const auto & a, const auto & b) { return a.idx < b.idx; });
-        const auto origin = *std::prev(pastOrigin);
-        const auto offset = offsets[idx];
-        return {offset.line, offset.column, origin.origin};
+            this is guaranteed to never rewind origin.begin because the first
+            key is always 0. */
+        const auto pastOrigin = origins.upper_bound(idx);
+        return &std::prev(pastOrigin)->second;
+    }
+
+public:
+    Origin addOrigin(Pos::Origin origin, size_t size)
+    {
+        uint32_t offset = 0;
+        if (auto it = origins.rbegin(); it != origins.rend())
+            offset = it->first + it->second.size;
+        // +1 because all PosIdx are offset by 1 to begin with, and
+        // another +1 to ensure that all origins can point to EOF, eg
+        // on (invalid) empty inputs.
+        if (2 + offset + size < offset)
+            return Origin{origin, offset, 0};
+        return origins.emplace(offset, Origin{origin, offset, size}).first->second;
+    }
+
+    PosIdx add(const Origin & origin, size_t offset)
+    {
+        if (offset > origin.size)
+            return PosIdx();
+        return PosIdx(1 + origin.offset + offset);
+    }
+
+    Pos operator[](PosIdx p) const;
+
+    Pos::Origin originOf(PosIdx p) const
+    {
+        if (auto o = resolve(p))
+            return o->origin;
+        return std::monostate{};
     }
 };