mtmd: add mtmd_bitmap_set_mergeable (#27348)

tests/test-mtmd-impl.cpp

@@ -1,9 +1,12 @@
 #include "testing.h"
 
 #include "mtmd-image.h"
+#include "mtmd-internal.h"
 
 #include <iostream>
+#include <stdexcept>
 #include <string>
+#include <tuple>
 #include <utility>
 #include <vector>
 
@@ -67,6 +70,73 @@ MAKE_TEST(test_image_preprocessor_lfm2) {
     }
 }
 
+//
+// mtmd temporal merge
+//
+
+MAKE_TEST(test_temporal_merge_grouping) {
+    std::vector<mtmd::bitmap_ptr> pool; // keeps the bitmaps alive until the end of the test
+
+    // spec chars:
+    //   v = video frame, w = video frame of another size, a = audio, i = plain image, t = text
+    auto make_parts = [&pool](const std::string & spec) {
+        std::vector<mtmd_input_part> parts;
+        for (char c : spec) {
+            if (c == 't') {
+                parts.push_back({ "hello", nullptr });
+                continue;
+            }
+            mtmd_bitmap * bm = nullptr;
+            switch (c) {
+                case 'v': bm = mtmd_bitmap_init(100, 100, nullptr);   break;
+                case 'w': bm = mtmd_bitmap_init(200, 200, nullptr);   break;
+                case 'a': bm = mtmd_bitmap_init_from_audio(100, nullptr); break;
+                case 'i': bm = mtmd_bitmap_init(100, 100, nullptr);   break;
+                default: throw std::runtime_error(std::string("unknown spec char: ") + c);
+            }
+            mtmd_bitmap_set_mergeable(bm, c != 'i');
+            pool.emplace_back(bm);
+            parts.push_back({ "", bm });
+        }
+        return parts;
+    };
+
+    // { parts, n_merge, expected size of each group }
+    const std::vector<std::tuple<std::string, int, std::string>> cases = {
+        { "vv",   2, "2"    },
+        { "vvv",  2, "21"   },
+        { "vvvv", 2, "22"   },
+        { "vvi",  2, "21"   },
+        { "tvvt", 2, "2"    },
+        { "vtv",  2, "11"   }, // text in between breaks the merge
+        { "vw",   2, "11"   }, // different sizes cannot be merged
+        { "aa",   2, "11"   }, // audio is never merged
+        { "ii",   2, "11"   }, // two unrelated images must stay separated
+        { "iv",   2, "11"   },
+        { "vi",   2, "11"   },
+        { "vv",   1, "11"   }, // model without temporal merge
+    };
+
+    for (const auto & [spec, n_merge, expected] : cases) {
+        auto parts  = make_parts(spec);
+        auto groups = mtmd_group_mergeable_bitmaps(parts, n_merge);
+
+        std::string actual;
+        for (const auto & group : groups) {
+            actual += std::to_string(group.size());
+        }
+
+        const std::string name = "\"" + spec + "\" with n_merge=" + std::to_string(n_merge);
+        t.assert_equal("groups for " + name, expected, actual);
+
+        size_t n_bitmap_parts = 0;
+        for (const auto & p : parts) {
+            n_bitmap_parts += p.bitmap != nullptr ? 1 : 0;
+        }
+        t.assert_equal("remaining bitmap parts for " + name, groups.size(), n_bitmap_parts);
+    }
+}
+
 //
 // main
 //