Remove the concept of "separate metadata"

This was an extra level of attribute indirection (features point
to metadata records which point to key and value strings) which was
intended to reduce the size of temporary storage for features with
large numbers of attributes that were also spread across large numbers
of tiles at maxzoom.

For other kinds of features, the extra indirection slowed things down
instead, and, especially when maxzoom guessing was being used, many more
features were having their metadata externalized than could actually
benefit from it.
This commit is contained in:
Erica Fischer
2022-12-29 14:47:27 -08:00
parent 3b2599f587
commit d05f8f8e4c
6 changed files with 40 additions and 208 deletions
+12 -23
View File
@@ -324,7 +324,7 @@ struct ordercmp {
}
} ordercmp;
void rewrite(drawvec &geom, int z, int nextzoom, int maxzoom, long long *bbox, unsigned tx, unsigned ty, int buffer, int *within, std::atomic<long long> *geompos, FILE **geomfile, const char *fname, signed char t, int layer, long long metastart, signed char feature_minzoom, int child_shards, int max_zoom_increment, long long seq, int tippecanoe_minzoom, int tippecanoe_maxzoom, int segment, unsigned *initial_x, unsigned *initial_y, std::vector<long long> &metakeys, std::vector<long long> &metavals, bool has_id, unsigned long long id, unsigned long long index, unsigned long long label_point, long long extent) {
void rewrite(drawvec &geom, int z, int nextzoom, int maxzoom, long long *bbox, unsigned tx, unsigned ty, int buffer, int *within, std::atomic<long long> *geompos, FILE **geomfile, const char *fname, signed char t, int layer, signed char feature_minzoom, int child_shards, int max_zoom_increment, long long seq, int tippecanoe_minzoom, int tippecanoe_maxzoom, int segment, unsigned *initial_x, unsigned *initial_y, std::vector<long long> &metakeys, std::vector<long long> &metavals, bool has_id, unsigned long long id, unsigned long long index, unsigned long long label_point, long long extent) {
if (geom.size() > 0 && (nextzoom <= maxzoom || additional[A_EXTEND_ZOOMS])) {
int xo, yo;
int span = 1 << (nextzoom - z);
@@ -411,18 +411,15 @@ void rewrite(drawvec &geom, int z, int nextzoom, int maxzoom, long long *bbox, u
sf.tippecanoe_minzoom = tippecanoe_minzoom;
sf.has_tippecanoe_maxzoom = tippecanoe_maxzoom != -1;
sf.tippecanoe_maxzoom = tippecanoe_maxzoom;
sf.metapos = metastart;
sf.geometry = geom2;
sf.index = index;
sf.label_point = label_point;
sf.extent = extent;
sf.feature_minzoom = feature_minzoom;
if (metastart < 0) {
for (size_t i = 0; i < metakeys.size(); i++) {
sf.keys.push_back(metakeys[i]);
sf.values.push_back(metavals[i]);
}
for (size_t i = 0; i < metakeys.size(); i++) {
sf.keys.push_back(metakeys[i]);
sf.values.push_back(metavals[i]);
}
serialize_feature(geomfile[j], &sf, &geompos[j], fname, SHIFT_RIGHT(initial_x[segment]), SHIFT_RIGHT(initial_y[segment]), true);
@@ -1283,7 +1280,6 @@ long long choose_minextent(std::vector<long long> &extents, double f) {
struct write_tile_args {
struct task *tasks = NULL;
char *metabase = NULL;
char *stringpool = NULL;
int min_detail = 0;
sqlite3 *outdb = NULL;
@@ -1306,7 +1302,6 @@ struct write_tile_args {
int low_detail = 0;
double simplification = 0;
std::atomic<long long> *most = NULL;
long long *meta_off = NULL;
long long *pool_off = NULL;
unsigned *initial_x = NULL;
unsigned *initial_y = NULL;
@@ -1429,9 +1424,9 @@ void remove_attributes(serial_feature &sf, std::set<std::string> const &exclude_
}
}
serial_feature next_feature(FILE *geoms, std::atomic<long long> *geompos_in, char *metabase, long long *meta_off, int z, unsigned tx, unsigned ty, unsigned *initial_x, unsigned *initial_y, long long *original_features, long long *unclipped_features, int nextzoom, int maxzoom, int minzoom, int max_zoom_increment, size_t pass, std::atomic<long long> *along, long long alongminus, int buffer, int *within, FILE **geomfile, std::atomic<long long> *geompos, std::atomic<double> *oprogress, double todo, const char *fname, int child_shards, struct json_object *filter, const char *stringpool, long long *pool_off, std::vector<std::vector<std::string>> *layer_unmaps, bool first_time) {
serial_feature next_feature(FILE *geoms, std::atomic<long long> *geompos_in, int z, unsigned tx, unsigned ty, unsigned *initial_x, unsigned *initial_y, long long *original_features, long long *unclipped_features, int nextzoom, int maxzoom, int minzoom, int max_zoom_increment, size_t pass, std::atomic<long long> *along, long long alongminus, int buffer, int *within, FILE **geomfile, std::atomic<long long> *geompos, std::atomic<double> *oprogress, double todo, const char *fname, int child_shards, struct json_object *filter, const char *stringpool, long long *pool_off, std::vector<std::vector<std::string>> *layer_unmaps, bool first_time) {
while (1) {
serial_feature sf = deserialize_feature(geoms, geompos_in, metabase, meta_off, z, tx, ty, initial_x, initial_y);
serial_feature sf = deserialize_feature(geoms, geompos_in, z, tx, ty, initial_x, initial_y);
if (sf.t < 0) {
return sf;
}
@@ -1460,7 +1455,7 @@ serial_feature next_feature(FILE *geoms, std::atomic<long long> *geompos_in, cha
if (first_time && pass == 0) { /* only write out the next zoom once, even if we retry */
if (sf.tippecanoe_maxzoom == -1 || sf.tippecanoe_maxzoom >= nextzoom) {
rewrite(sf.geometry, z, nextzoom, maxzoom, sf.bbox, tx, ty, buffer, within, geompos, geomfile, fname, sf.t, sf.layer, sf.metapos, sf.feature_minzoom, child_shards, max_zoom_increment, sf.seq, sf.tippecanoe_minzoom, sf.tippecanoe_maxzoom, sf.segment, initial_x, initial_y, sf.keys, sf.values, sf.has_id, sf.id, sf.index, sf.label_point, sf.extent);
rewrite(sf.geometry, z, nextzoom, maxzoom, sf.bbox, tx, ty, buffer, within, geompos, geomfile, fname, sf.t, sf.layer, sf.feature_minzoom, child_shards, max_zoom_increment, sf.seq, sf.tippecanoe_minzoom, sf.tippecanoe_maxzoom, sf.segment, initial_x, initial_y, sf.keys, sf.values, sf.has_id, sf.id, sf.index, sf.label_point, sf.extent);
}
}
@@ -1563,8 +1558,6 @@ serial_feature next_feature(FILE *geoms, std::atomic<long long> *geompos_in, cha
struct run_prefilter_args {
FILE *geoms = NULL;
std::atomic<long long> *geompos_in = NULL;
char *metabase = NULL;
long long *meta_off = NULL;
int z = 0;
unsigned tx = 0;
unsigned ty = 0;
@@ -1600,7 +1593,7 @@ void *run_prefilter(void *v) {
json_writer state(rpa->prefilter_fp);
while (1) {
serial_feature sf = next_feature(rpa->geoms, rpa->geompos_in, rpa->metabase, rpa->meta_off, rpa->z, rpa->tx, rpa->ty, rpa->initial_x, rpa->initial_y, rpa->original_features, rpa->unclipped_features, rpa->nextzoom, rpa->maxzoom, rpa->minzoom, rpa->max_zoom_increment, rpa->pass, rpa->along, rpa->alongminus, rpa->buffer, rpa->within, rpa->geomfile, rpa->geompos, rpa->oprogress, rpa->todo, rpa->fname, rpa->child_shards, rpa->filter, rpa->stringpool, rpa->pool_off, rpa->layer_unmaps, rpa->first_time);
serial_feature sf = next_feature(rpa->geoms, rpa->geompos_in, rpa->z, rpa->tx, rpa->ty, rpa->initial_x, rpa->initial_y, rpa->original_features, rpa->unclipped_features, rpa->nextzoom, rpa->maxzoom, rpa->minzoom, rpa->max_zoom_increment, rpa->pass, rpa->along, rpa->alongminus, rpa->buffer, rpa->within, rpa->geomfile, rpa->geompos, rpa->oprogress, rpa->todo, rpa->fname, rpa->child_shards, rpa->filter, rpa->stringpool, rpa->pool_off, rpa->layer_unmaps, rpa->first_time);
if (sf.t < 0) {
break;
}
@@ -1850,7 +1843,7 @@ void add_sample_to(std::vector<T> &vals, T val, size_t &increment, size_t seq) {
}
}
long long write_tile(FILE *geoms, std::atomic<long long> *geompos_in, char *metabase, char *stringpool, int z, const unsigned tx, const unsigned ty, const int detail, int min_detail, sqlite3 *outdb, const char *outdir, int buffer, const char *fname, FILE **geomfile, int minzoom, int maxzoom, double todo, std::atomic<long long> *along, long long alongminus, double gamma, int child_shards, long long *meta_off, long long *pool_off, unsigned *initial_x, unsigned *initial_y, std::atomic<int> *running, double simplification, std::vector<std::map<std::string, layermap_entry>> *layermaps, std::vector<std::vector<std::string>> *layer_unmaps, size_t tiling_seg, size_t pass, unsigned long long mingap, long long minextent, double fraction, const char *prefilter, const char *postfilter, struct json_object *filter, write_tile_args *arg, atomic_strategy *strategy) {
long long write_tile(FILE *geoms, std::atomic<long long> *geompos_in, char *stringpool, int z, const unsigned tx, const unsigned ty, const int detail, int min_detail, sqlite3 *outdb, const char *outdir, int buffer, const char *fname, FILE **geomfile, int minzoom, int maxzoom, double todo, std::atomic<long long> *along, long long alongminus, double gamma, int child_shards, long long *pool_off, unsigned *initial_x, unsigned *initial_y, std::atomic<int> *running, double simplification, std::vector<std::map<std::string, layermap_entry>> *layermaps, std::vector<std::vector<std::string>> *layer_unmaps, size_t tiling_seg, size_t pass, unsigned long long mingap, long long minextent, double fraction, const char *prefilter, const char *postfilter, struct json_object *filter, write_tile_args *arg, atomic_strategy *strategy) {
double merge_fraction = 1;
double mingap_fraction = 1;
double minextent_fraction = 1;
@@ -1954,8 +1947,6 @@ long long write_tile(FILE *geoms, std::atomic<long long> *geompos_in, char *meta
rpa.geoms = geoms;
rpa.geompos_in = geompos_in;
rpa.metabase = metabase;
rpa.meta_off = meta_off;
rpa.z = z;
rpa.tx = tx;
rpa.ty = ty;
@@ -2003,7 +1994,7 @@ long long write_tile(FILE *geoms, std::atomic<long long> *geompos_in, char *meta
ssize_t which_partial = -1;
if (prefilter == NULL) {
sf = next_feature(geoms, geompos_in, metabase, meta_off, z, tx, ty, initial_x, initial_y, &original_features, &unclipped_features, nextzoom, maxzoom, minzoom, max_zoom_increment, pass, along, alongminus, buffer, within, geomfile, geompos, &oprogress, todo, fname, child_shards, filter, stringpool, pool_off, layer_unmaps, first_time);
sf = next_feature(geoms, geompos_in, z, tx, ty, initial_x, initial_y, &original_features, &unclipped_features, nextzoom, maxzoom, minzoom, max_zoom_increment, pass, along, alongminus, buffer, within, geomfile, geompos, &oprogress, todo, fname, child_shards, filter, stringpool, pool_off, layer_unmaps, first_time);
} else {
sf = parse_feature(prefilter_jp, z, tx, ty, layermaps, tiling_seg, layer_unmaps, postfilter != NULL);
}
@@ -2837,7 +2828,7 @@ void *run_thread(void *vargs) {
// fprintf(stderr, "%d/%u/%u\n", z, x, y);
long long len = write_tile(geom, &geompos, arg->metabase, arg->stringpool, z, x, y, z == arg->maxzoom ? arg->full_detail : arg->low_detail, arg->min_detail, arg->outdb, arg->outdir, arg->buffer, arg->fname, arg->geomfile, arg->minzoom, arg->maxzoom, arg->todo, arg->along, geompos, arg->gamma, arg->child_shards, arg->meta_off, arg->pool_off, arg->initial_x, arg->initial_y, arg->running, arg->simplification, arg->layermaps, arg->layer_unmaps, arg->tiling_seg, arg->pass, arg->mingap, arg->minextent, arg->fraction, arg->prefilter, arg->postfilter, arg->filter, arg, arg->strategy);
long long len = write_tile(geom, &geompos, arg->stringpool, z, x, y, z == arg->maxzoom ? arg->full_detail : arg->low_detail, arg->min_detail, arg->outdb, arg->outdir, arg->buffer, arg->fname, arg->geomfile, arg->minzoom, arg->maxzoom, arg->todo, arg->along, geompos, arg->gamma, arg->child_shards, arg->pool_off, arg->initial_x, arg->initial_y, arg->running, arg->simplification, arg->layermaps, arg->layer_unmaps, arg->tiling_seg, arg->pass, arg->mingap, arg->minextent, arg->fraction, arg->prefilter, arg->postfilter, arg->filter, arg, arg->strategy);
if (len < 0) {
int *err = &arg->err;
@@ -2902,7 +2893,7 @@ void *run_thread(void *vargs) {
return NULL;
}
int traverse_zooms(int *geomfd, off_t *geom_size, char *metabase, char *stringpool, std::atomic<unsigned> *midx, std::atomic<unsigned> *midy, int &maxzoom, int minzoom, sqlite3 *outdb, const char *outdir, int buffer, const char *fname, const char *tmpdir, double gamma, int full_detail, int low_detail, int min_detail, long long *meta_off, long long *pool_off, unsigned *initial_x, unsigned *initial_y, double simplification, double maxzoom_simplification, std::vector<std::map<std::string, layermap_entry>> &layermaps, const char *prefilter, const char *postfilter, std::map<std::string, attribute_op> const *attribute_accum, struct json_object *filter, std::vector<strategy> &strategies) {
int traverse_zooms(int *geomfd, off_t *geom_size, char *stringpool, std::atomic<unsigned> *midx, std::atomic<unsigned> *midy, int &maxzoom, int minzoom, sqlite3 *outdb, const char *outdir, int buffer, const char *fname, const char *tmpdir, double gamma, int full_detail, int low_detail, int min_detail, long long *pool_off, unsigned *initial_x, unsigned *initial_y, double simplification, double maxzoom_simplification, std::vector<std::map<std::string, layermap_entry>> &layermaps, const char *prefilter, const char *postfilter, std::map<std::string, attribute_op> const *attribute_accum, struct json_object *filter, std::vector<strategy> &strategies) {
last_progress = 0;
// The existing layermaps are one table per input thread.
@@ -3049,7 +3040,6 @@ int traverse_zooms(int *geomfd, off_t *geom_size, char *metabase, char *stringpo
atomic_strategy strategy;
for (size_t thread = 0; thread < threads; thread++) {
args[thread].metabase = metabase;
args[thread].stringpool = stringpool;
args[thread].min_detail = min_detail;
args[thread].outdb = outdb; // locked with db_lock
@@ -3086,7 +3076,6 @@ int traverse_zooms(int *geomfd, off_t *geom_size, char *metabase, char *stringpo
args[thread].full_detail = full_detail;
args[thread].low_detail = low_detail;
args[thread].most = &most; // locked with var_lock
args[thread].meta_off = meta_off;
args[thread].pool_off = pool_off;
args[thread].initial_x = initial_x;
args[thread].initial_y = initial_y;