[PATCH v4 14/14] box: specify indexes in user-friendly form

[Kirill Shcherbatov <kshcherbatov@tarantool.org>]

Since now it is possible to create indexes by JSON-path
using field names specified in format.

Closes #1012.

@TarantoolBot document
Title: Indexes by JSON path
Sometimes field data could have complex document structure.
When this structure is consistent across whole document,
you are able to create an index by JSON path.

Example:
s:create_index('json_index', {parts = {{'FIO["fname"]', 'str'}}})
---
 src/box/lua/index.c        | 60 ++++++++++++++++++++++++++++++++++++++++++++++
 src/box/lua/schema.lua     | 22 ++++++++---------
 test/engine/tuple.result   | 41 +++++++++++++++++++++++++++++++
 test/engine/tuple.test.lua | 12 ++++++++++
 4 files changed, 124 insertions(+), 11 deletions(-)

diff --git a/src/box/lua/index.c b/src/box/lua/index.c
index ef89c39..21c299d 100644
--- a/src/box/lua/index.c
+++ b/src/box/lua/index.c
@@ -35,6 +35,9 @@
 #include "box/info.h"
 #include "box/lua/info.h"
 #include "box/lua/tuple.h"
+#include "box/schema.h"
+#include "box/tuple_format.h"
+#include "json/path.h"
 #include "box/lua/misc.h" /* lbox_encode_tuple_on_gc() */
 
 /** {{{ box.index Lua library: access to spaces and indexes
@@ -328,6 +331,62 @@ lbox_index_compact(lua_State *L)
 	return 0;
 }
 
+static int
+lbox_index_resolve_path(struct lua_State *L)
+{
+	if (lua_gettop(L) != 3 ||
+	    !lua_isnumber(L, 1) || !lua_isnumber(L, 2) || !lua_isstring(L, 3)) {
+		return luaL_error(L, "Usage box.internal."
+				     "path_resolve(part_id, space_id, path)");
+	}
+	uint32_t part_id = lua_tonumber(L, 1);
+	uint32_t space_id = lua_tonumber(L, 2);
+	size_t path_len;
+	const char *path = lua_tolstring(L, 3, &path_len);
+	struct space *space = space_cache_find(space_id);
+	if (space == NULL)
+		return luaT_error(L);
+	struct json_path_parser parser;
+	struct json_path_node node;
+	json_path_parser_create(&parser, path, path_len);
+	int rc = json_path_next(&parser, &node);
+	if (rc != 0) {
+		const char *err_msg =
+			tt_sprintf("options.parts[%d]: error in path on "
+				   "position %d", part_id, rc);
+		diag_set(ClientError, ER_ILLEGAL_PARAMS, err_msg);
+		return luaT_error(L);
+	}
+	assert(space->format != NULL && space->format->dict != NULL);
+	uint32_t fieldno;
+	if (node.type == JSON_PATH_NUM &&
+	    (fieldno = node.num - TUPLE_INDEX_BASE) >=
+	    space->format->field_count) {
+		const char *err_msg =
+			tt_sprintf("options.parts[%d]: field '%d' referenced "
+				   "in path is greater than format field "
+				   "count %d", part_id,
+				   fieldno + TUPLE_INDEX_BASE,
+				   space->format->field_count);
+		diag_set(ClientError, ER_ILLEGAL_PARAMS, err_msg);
+		return luaT_error(L);
+	} else if (node.type == JSON_PATH_STR &&
+		   tuple_fieldno_by_name(space->format->dict, node.str, node.len,
+					 field_name_hash(node.str, node.len),
+					 &fieldno) != 0) {
+		const char *err_msg =
+			tt_sprintf("options.parts[%d]: field was not found by "
+				   "name '%.*s'", part_id, node.len, node.str);
+		diag_set(ClientError, ER_ILLEGAL_PARAMS, err_msg);
+		return luaT_error(L);
+	}
+	fieldno += TUPLE_INDEX_BASE;
+	path += parser.offset;
+	lua_pushnumber(L, fieldno);
+	lua_pushstring(L, path);
+	return 2;
+}
+
 /* }}} */
 
 void
@@ -365,6 +424,7 @@ box_lua_index_init(struct lua_State *L)
 		{"truncate", lbox_truncate},
 		{"stat", lbox_index_stat},
 		{"compact", lbox_index_compact},
+		{"path_resolve", lbox_index_resolve_path},
 		{NULL, NULL}
 	};
 
diff --git a/src/box/lua/schema.lua b/src/box/lua/schema.lua
index 540a2a5..8fca3bd 100644
--- a/src/box/lua/schema.lua
+++ b/src/box/lua/schema.lua
@@ -556,7 +556,7 @@ local function update_index_parts_1_6_0(parts)
     return result
 end
 
-local function update_index_parts(format, parts)
+local function update_index_parts(format, parts, space_id)
     if type(parts) ~= "table" then
         box.error(box.error.ILLEGAL_PARAMS,
         "options.parts parameter should be a table")
@@ -607,16 +607,16 @@ local function update_index_parts(format, parts)
             box.error(box.error.ILLEGAL_PARAMS,
                       "options.parts[" .. i .. "]: field (name or number) is expected")
         elseif type(part.field) == 'string' then
-            for k,v in pairs(format) do
-                if v.name == part.field then
-                    part.field = k
-                    break
-                end
-            end
-            if type(part.field) == 'string' then
+            local idx, path = box.internal.path_resolve(i, space_id, part.field)
+            if part.path ~= nil and part.path ~= path then
                 box.error(box.error.ILLEGAL_PARAMS,
-                          "options.parts[" .. i .. "]: field was not found by name '" .. part.field .. "'")
+                          "options.parts[" .. i .. "]: field path '"..
+                          part.path.." doesn't math path resolved by name '" ..
+                          part.field .. "'")
             end
+            parts_can_be_simplified = parts_can_be_simplified and path == nil
+            part.field = idx
+            part.path = path or part.path
         elseif part.field == 0 then
             box.error(box.error.ILLEGAL_PARAMS,
                       "options.parts[" .. i .. "]: field (number) must be one-based")
@@ -767,7 +767,7 @@ box.schema.index.create = function(space_id, name, options)
         end
     end
     local parts, parts_can_be_simplified =
-        update_index_parts(format, options.parts)
+        update_index_parts(format, options.parts, space_id)
     -- create_index() options contains type, parts, etc,
     -- stored separately. Remove these members from index_opts
     local index_opts = {
@@ -934,7 +934,7 @@ box.schema.index.alter = function(space_id, index_id, options)
     if options.parts then
         local parts_can_be_simplified
         parts, parts_can_be_simplified =
-            update_index_parts(format, options.parts)
+            update_index_parts(format, options.parts, space_id)
         -- save parts in old format if possible
         if parts_can_be_simplified then
             parts = simplify_index_parts(parts)
diff --git a/test/engine/tuple.result b/test/engine/tuple.result
index 1842420..4dd34fd 100644
--- a/test/engine/tuple.result
+++ b/test/engine/tuple.result
@@ -1002,6 +1002,47 @@ assert(idx ~= nil)
 ---
 - true
 ...
+format = {{'int1', 'unsigned'}, {'int2', 'unsigned'}, {'data', 'array'}, {'int3', 'unsigned'}, {'int4', 'unsigned'}}
+---
+...
+s:format(format)
+---
+- error: Field 3 has type 'map' in one index, but type 'array' in another
+...
+format = {{'int1', 'unsigned'}, {'int2', 'unsigned'}, {'data', 'map'}, {'int3', 'unsigned'}, {'int4', 'unsigned'}}
+---
+...
+s:format(format)
+---
+...
+s:create_index('test3', {parts = {{2, 'number'}, {']sad.FIO["fname"]', 'str'}}})
+---
+- error: 'Illegal parameters, options.parts[2]: error in path on position 1'
+...
+s:create_index('test3', {parts = {{2, 'number'}, {'[666].FIO["fname"]', 'str'}}})
+---
+- error: 'Illegal parameters, options.parts[2]: field ''666'' referenced in path is
+    greater than format field count 5'
+...
+s:create_index('test3', {parts = {{2, 'number'}, {'invalid.FIO["fname"]', 'str'}}})
+---
+- error: 'Illegal parameters, options.parts[2]: field was not found by name ''invalid'''
+...
+idx3 = s:create_index('test3', {parts = {{2, 'number'}, {'data.FIO["fname"]', 'str'}}})
+---
+...
+assert(idx3 ~= nil)
+---
+- true
+...
+assert(idx3.parts[2].path == "[\"FIO\"][\"fname\"]")
+---
+- true
+...
+-- Vinyl has optimizations that omit index checks, so errors could differ.
+idx3:drop()
+---
+...
 s:insert{7, 7, {town = 'London', FIO = 666}, 4, 5}
 ---
 - error: 'Tuple field 3 type does not match one required by operation: expected map'
diff --git a/test/engine/tuple.test.lua b/test/engine/tuple.test.lua
index 8be6505..f8d99a0 100644
--- a/test/engine/tuple.test.lua
+++ b/test/engine/tuple.test.lua
@@ -326,6 +326,18 @@ s:create_index('test1', {parts = {{2, 'number'}, {3, 'str', path = '[1].sname'},
 s:create_index('test1', {parts = {{2, 'number'}, {3, 'str', path = 'FIO....fname'}}})
 idx = s:create_index('test1', {parts = {{2, 'number'}, {3, 'str', path = '["FIO"]["fname"]'}, {3, 'str', path = '["FIO"]["sname"]'}}})
 assert(idx ~= nil)
+format = {{'int1', 'unsigned'}, {'int2', 'unsigned'}, {'data', 'array'}, {'int3', 'unsigned'}, {'int4', 'unsigned'}}
+s:format(format)
+format = {{'int1', 'unsigned'}, {'int2', 'unsigned'}, {'data', 'map'}, {'int3', 'unsigned'}, {'int4', 'unsigned'}}
+s:format(format)
+s:create_index('test3', {parts = {{2, 'number'}, {']sad.FIO["fname"]', 'str'}}})
+s:create_index('test3', {parts = {{2, 'number'}, {'[666].FIO["fname"]', 'str'}}})
+s:create_index('test3', {parts = {{2, 'number'}, {'invalid.FIO["fname"]', 'str'}}})
+idx3 = s:create_index('test3', {parts = {{2, 'number'}, {'data.FIO["fname"]', 'str'}}})
+assert(idx3 ~= nil)
+assert(idx3.parts[2].path == "[\"FIO\"][\"fname\"]")
+-- Vinyl has optimizations that omit index checks, so errors could differ.
+idx3:drop()
 s:insert{7, 7, {town = 'London', FIO = 666}, 4, 5}
 s:insert{7, 7, {town = 'London', FIO = {fname = 666, sname = 'Bond'}}, 4, 5}
 s:insert{7, 7, {town = 'London', FIO = {fname = "James"}}, 4, 5}
-- 
2.7.4