1/*
2 * Copyright (C) 2016 The Android Open Source Project
3 *
4 * Licensed under the Apache License, Version 2.0 (the "License");
5 * you may not use this file except in compliance with the License.
6 * You may obtain a copy of the License at
7 *
8 *      http://www.apache.org/licenses/LICENSE-2.0
9 *
10 * Unless required by applicable law or agreed to in writing, software
11 * distributed under the License is distributed on an "AS IS" BASIS,
12 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 * See the License for the specific language governing permissions and
14 * limitations under the License.
15 */
16
17#include <cstdint>
18#include <cstdlib>
19#include <cstring>
20#include <string>
21#include <unordered_map>
22#include <unordered_set>
23
24#include <androidfw/LocaleData.h>
25
26namespace android {
27
28#include "LocaleDataTables.cpp"
29
30inline uint32_t packLocale(const char* language, const char* region) {
31    return (((uint8_t) language[0]) << 24u) | (((uint8_t) language[1]) << 16u) |
32           (((uint8_t) region[0]) << 8u) | ((uint8_t) region[1]);
33}
34
35inline uint32_t dropRegion(uint32_t packed_locale) {
36    return packed_locale & 0xFFFF0000lu;
37}
38
39inline bool hasRegion(uint32_t packed_locale) {
40    return (packed_locale & 0x0000FFFFlu) != 0;
41}
42
43const size_t SCRIPT_LENGTH = 4;
44const size_t SCRIPT_PARENTS_COUNT = sizeof(SCRIPT_PARENTS)/sizeof(SCRIPT_PARENTS[0]);
45const uint32_t PACKED_ROOT = 0; // to represent the root locale
46
47uint32_t findParent(uint32_t packed_locale, const char* script) {
48    if (hasRegion(packed_locale)) {
49        for (size_t i = 0; i < SCRIPT_PARENTS_COUNT; i++) {
50            if (memcmp(script, SCRIPT_PARENTS[i].script, SCRIPT_LENGTH) == 0) {
51                auto map = SCRIPT_PARENTS[i].map;
52                auto lookup_result = map->find(packed_locale);
53                if (lookup_result != map->end()) {
54                    return lookup_result->second;
55                }
56                break;
57            }
58        }
59        return dropRegion(packed_locale);
60    }
61    return PACKED_ROOT;
62}
63
64// Find the ancestors of a locale, and fill 'out' with it (assumes out has enough
65// space). If any of the members of stop_list was seen, write it in the
66// output but stop afterwards.
67//
68// This also outputs the index of the last written ancestor in the stop_list
69// to stop_list_index, which will be -1 if it is not found in the stop_list.
70//
71// Returns the number of ancestors written in the output, which is always
72// at least one.
73//
74// (If 'out' is nullptr, we do everything the same way but we simply don't write
75// any results in 'out'.)
76size_t findAncestors(uint32_t* out, ssize_t* stop_list_index,
77                     uint32_t packed_locale, const char* script,
78                     const uint32_t* stop_list, size_t stop_set_length) {
79    uint32_t ancestor = packed_locale;
80    size_t count = 0;
81    do {
82        if (out != nullptr) out[count] = ancestor;
83        count++;
84        for (size_t i = 0; i < stop_set_length; i++) {
85            if (stop_list[i] == ancestor) {
86                *stop_list_index = (ssize_t) i;
87                return count;
88            }
89        }
90        ancestor = findParent(ancestor, script);
91    } while (ancestor != PACKED_ROOT);
92    *stop_list_index = (ssize_t) -1;
93    return count;
94}
95
96size_t findDistance(uint32_t supported,
97                    const char* script,
98                    const uint32_t* request_ancestors,
99                    size_t request_ancestors_count) {
100    ssize_t request_ancestors_index;
101    const size_t supported_ancestor_count = findAncestors(
102            nullptr, &request_ancestors_index,
103            supported, script,
104            request_ancestors, request_ancestors_count);
105    // Since both locales share the same root, there will always be a shared
106    // ancestor, so the distance in the parent tree is the sum of the distance
107    // of 'supported' to the lowest common ancestor (number of ancestors
108    // written for 'supported' minus 1) plus the distance of 'request' to the
109    // lowest common ancestor (the index of the ancestor in request_ancestors).
110    return supported_ancestor_count + request_ancestors_index - 1;
111}
112
113inline bool isRepresentative(uint32_t language_and_region, const char* script) {
114    const uint64_t packed_locale = (
115            (((uint64_t) language_and_region) << 32u) |
116            (((uint64_t) script[0]) << 24u) |
117            (((uint64_t) script[1]) << 16u) |
118            (((uint64_t) script[2]) <<  8u) |
119            ((uint64_t) script[3]));
120
121    return (REPRESENTATIVE_LOCALES.count(packed_locale) != 0);
122}
123
124int localeDataCompareRegions(
125        const char* left_region, const char* right_region,
126        const char* requested_language, const char* requested_script,
127        const char* requested_region) {
128
129    if (left_region[0] == right_region[0] && left_region[1] == right_region[1]) {
130        return 0;
131    }
132    const uint32_t left = packLocale(requested_language, left_region);
133    const uint32_t right = packLocale(requested_language, right_region);
134    const uint32_t request = packLocale(requested_language, requested_region);
135
136    uint32_t request_ancestors[MAX_PARENT_DEPTH+1];
137    ssize_t left_right_index;
138    // Find the parents of the request, but stop as soon as we saw left or right
139    const uint32_t left_and_right[] = {left, right};
140    const size_t ancestor_count = findAncestors(
141            request_ancestors, &left_right_index,
142            request, requested_script,
143            left_and_right, sizeof(left_and_right)/sizeof(left_and_right[0]));
144    if (left_right_index == 0) { // We saw left earlier
145        return 1;
146    }
147    if (left_right_index == 1) { // We saw right earlier
148        return -1;
149    }
150
151    // If we are here, neither left nor right are an ancestor of the
152    // request. This means that all the ancestors have been computed and
153    // the last ancestor is just the language by itself. We will use the
154    // distance in the parent tree for determining the better match.
155    const size_t left_distance = findDistance(
156            left, requested_script, request_ancestors, ancestor_count);
157    const size_t right_distance = findDistance(
158            right, requested_script, request_ancestors, ancestor_count);
159    if (left_distance != right_distance) {
160        return (int) right_distance - (int) left_distance; // smaller distance is better
161    }
162
163    // If we are here, left and right are equidistant from the request. We will
164    // try and see if any of them is a representative locale.
165    const bool left_is_representative = isRepresentative(left, requested_script);
166    const bool right_is_representative = isRepresentative(right, requested_script);
167    if (left_is_representative != right_is_representative) {
168        return (int) left_is_representative - (int) right_is_representative;
169    }
170
171    // We have no way of figuring out which locale is a better match. For
172    // the sake of stability, we consider the locale with the lower region
173    // code (in dictionary order) better, with two-letter codes before
174    // three-digit codes (since two-letter codes are more specific).
175    return (int64_t) right - (int64_t) left;
176}
177
178void localeDataComputeScript(char out[4], const char* language, const char* region) {
179    if (language[0] == '\0') {
180        memset(out, '\0', SCRIPT_LENGTH);
181        return;
182    }
183    uint32_t lookup_key = packLocale(language, region);
184    auto lookup_result = LIKELY_SCRIPTS.find(lookup_key);
185    if (lookup_result == LIKELY_SCRIPTS.end()) {
186        // We couldn't find the locale. Let's try without the region
187        if (region[0] != '\0') {
188            lookup_key = dropRegion(lookup_key);
189            lookup_result = LIKELY_SCRIPTS.find(lookup_key);
190            if (lookup_result != LIKELY_SCRIPTS.end()) {
191                memcpy(out, SCRIPT_CODES[lookup_result->second], SCRIPT_LENGTH);
192                return;
193            }
194        }
195        // We don't know anything about the locale
196        memset(out, '\0', SCRIPT_LENGTH);
197        return;
198    } else {
199        // We found the locale.
200        memcpy(out, SCRIPT_CODES[lookup_result->second], SCRIPT_LENGTH);
201    }
202}
203
204const uint32_t ENGLISH_STOP_LIST[2] = {
205    0x656E0000lu, // en
206    0x656E8400lu, // en-001
207};
208const char ENGLISH_CHARS[2] = {'e', 'n'};
209const char LATIN_CHARS[4] = {'L', 'a', 't', 'n'};
210
211bool localeDataIsCloseToUsEnglish(const char* region) {
212    const uint32_t locale = packLocale(ENGLISH_CHARS, region);
213    ssize_t stop_list_index;
214    findAncestors(nullptr, &stop_list_index, locale, LATIN_CHARS, ENGLISH_STOP_LIST, 2);
215    // A locale is like US English if we see "en" before "en-001" in its ancestor list.
216    return stop_list_index == 0; // 'en' is first in ENGLISH_STOP_LIST
217}
218
219} // namespace android
220