Compare commits

...
5 Commits
10 changed files with 24146 additions and 28547 deletions
+20 -20
View File
@@ -527,22 +527,22 @@ accumulate-test:
# and 99 without it
test `grep '"POP1950": null' tests/ne_110m_populated_places_nulls/in.json | wc -l` == 99
./tippecanoe -yNAME -yPOP1950 -yclustered:cluster_size -yclustered:unrelated -q -z3 -r1.75 -b0 -f -e tests/pbf/accum.dir --accumulate-numeric-attributes=clustered --set-attribute '{"clustered:cluster_size":1}' --accumulate-attribute '{"clustered:cluster_size":"sum"}' --retain-points-multiplier 3 tests/ne_110m_populated_places_nulls/in.json
# at this drop rate, there are 6 points at z0 that have no POP1950s clustered onto them....
test `./tippecanoe-decode -c tests/pbf/accum.dir/0/0/0.pbf 0 0 0 | grep -v 'clustered:count:POP1950' | wc -l` == 78
# 35 of which have no POP1950 at all
test `./tippecanoe-decode -c tests/pbf/accum.dir/0/0/0.pbf 0 0 0 | grep -v 'POP1950' | wc -l` == 35
# 43 of which do have POP1950
test `./tippecanoe-decode -c tests/pbf/accum.dir/0/0/0.pbf 0 0 0 | grep -v 'clustered:count:POP1950' | grep 'POP1950' | wc -l` == 43
# plus 59 that are clustered
test `./tippecanoe-decode -c tests/pbf/accum.dir/0/0/0.pbf 0 0 0 | grep 'clustered:count:POP1950' | wc -l` == 59
# the 59 clustered POP1950s have a total count of 101
test `./tippecanoe-decode -c tests/pbf/accum.dir/0/0/0.pbf 0 0 0 | grep 'clustered:count:POP1950' | sed 's/.*"clustered:count:POP1950": //' | awk '{sum += $$1} END {print sum}'` == 101
# we have already established that there are 43 bare POP1950s
# at this drop rate, there are 61 points at z0 that have no POP1950s clustered onto them....
test `./tippecanoe-decode -c tests/pbf/accum.dir/0/0/0.pbf 0 0 0 | grep -v 'clustered:count:POP1950' | wc -l` == 61
# 26 of which have no POP1950 at all
test `./tippecanoe-decode -c tests/pbf/accum.dir/0/0/0.pbf 0 0 0 | grep -v 'POP1950' | wc -l` == 26
# 35 of which do have POP1950
test `./tippecanoe-decode -c tests/pbf/accum.dir/0/0/0.pbf 0 0 0 | grep -v 'clustered:count:POP1950' | grep 'POP1950' | wc -l` == 35
# plus 60 that are clustered
test `./tippecanoe-decode -c tests/pbf/accum.dir/0/0/0.pbf 0 0 0 | grep 'clustered:count:POP1950' | wc -l` == 60
# the 60 clustered POP1950s have a total count of 109
test `./tippecanoe-decode -c tests/pbf/accum.dir/0/0/0.pbf 0 0 0 | grep 'clustered:count:POP1950' | sed 's/.*"clustered:count:POP1950": //' | awk '{sum += $$1} END {print sum}'` == 109
# we have already established that there are 36 bare POP1950s
# which makes a total of 144, which is the total count expected
#
# meanwhile, regular attribute accumulation.
# there are 137 features in the z0 tile, and they all have clustered:cluster_size
test `./tippecanoe-decode -c tests/pbf/accum.dir/0/0/0.pbf 0 0 0 | grep 'clustered:cluster_size' | wc -l` == 137
# there are 121 features in the z0 tile, and they all have clustered:cluster_size
test `./tippecanoe-decode -c tests/pbf/accum.dir/0/0/0.pbf 0 0 0 | grep 'clustered:cluster_size' | wc -l` == 121
# there are no features that lack it.
test `./tippecanoe-decode -c tests/pbf/accum.dir/0/0/0.pbf 0 0 0 | grep -v 'clustered:cluster_size' | wc -l` == 0
# they add up to the 243 original features
@@ -550,15 +550,15 @@ accumulate-test:
# Make sure we do *not* accumulate a numeric attribute that already has the magic prefix:
test `./tippecanoe-decode -c tests/pbf/accum.dir/0/0/0.pbf 0 0 0 | grep sum:clustered:unrelated | wc -l` == 0
# But that we *do* preserve those attributes into the output features:
test `./tippecanoe-decode -c tests/pbf/accum.dir/0/0/0.pbf 0 0 0 | grep clustered:unrelated | wc -l` == 66
test `./tippecanoe-decode -c tests/pbf/accum.dir/0/0/0.pbf 0 0 0 | grep clustered:unrelated | wc -l` == 61
#
# on to the sums:
# in the original data set, the POP1950s that are present add up to 161590
test `grep '"POP1950": [0-9]' tests/ne_110m_populated_places_nulls/in.json | sed 's/.*"POP1950": //' | awk '{sum += $$1} END {print sum}' ` == 161590
# in the z0 tile, the clustered POP1950s add up to 113357
test `./tippecanoe-decode -c tests/pbf/accum.dir/0/0/0.pbf 0 0 0 | grep 'clustered:sum:POP1950' | sed 's/.*"clustered:sum:POP1950": //' | awk '{sum += $$1} END {print sum}'` == 113357
# and the non-clustered ones add up to 48233
test `./tippecanoe-decode -c tests/pbf/accum.dir/0/0/0.pbf 0 0 0 | grep -v 'clustered:sum:POP1950' | grep POP1950 | sed 's/.*"POP1950": //' | awk '{sum += $$1} END {print sum}'` == 48233
# in the z0 tile, the clustered POP1950s add up to 116967
test `./tippecanoe-decode -c tests/pbf/accum.dir/0/0/0.pbf 0 0 0 | grep 'clustered:sum:POP1950' | sed 's/.*"clustered:sum:POP1950": //' | awk '{sum += $$1} END {print sum}'` == 116967
# and the non-clustered ones add up to 44623
test `./tippecanoe-decode -c tests/pbf/accum.dir/0/0/0.pbf 0 0 0 | grep -v 'clustered:sum:POP1950' | grep POP1950 | sed 's/.*"POP1950": //' | awk '{sum += $$1} END {print sum}'` == 44623
# which is the correct 161590
#
# OK, so do these still hold after megatile filtering?
@@ -599,11 +599,11 @@ accumulate-test:
# Now on to binning!
./tippecanoe-overzoom --assign-to-bins tests/pbf/h3-0-0-0.geojson --accumulate-numeric-attributes=clustered --accumulate-attribute '{"clustered:cluster_size":"sum"}' -o tests/pbf/bins-0-0-0.pbf tests/pbf/accum.dir/0/0/0.pbf 0/0/0 0/0/0
# Now there are 30 bins with POP1950 clusters
test `./tippecanoe-decode -c tests/pbf/bins-0-0-0.pbf 0 0 0 | grep 'clustered:count:POP1950' | wc -l` == 44
test `./tippecanoe-decode -c tests/pbf/bins-0-0-0.pbf 0 0 0 | grep 'clustered:count:POP1950' | wc -l` == 41
# There are none with bare POP1950 (which is expected; we should only have summary statistics)
test `./tippecanoe-decode -c tests/pbf/bins-0-0-0.pbf 0 0 0 | grep -v 'clustered:count:POP1950' | grep 'POP1950' | wc -l` == 0
# And 4 with no POP1950 at all
test `./tippecanoe-decode -c tests/pbf/bins-0-0-0.pbf 0 0 0 | grep -v 'POP1950' | wc -l` == 4
test `./tippecanoe-decode -c tests/pbf/bins-0-0-0.pbf 0 0 0 | grep -v 'POP1950' | wc -l` == 3
#
# the clustered and megatile-filtered and binned POP1950s add up to 161590
test `./tippecanoe-decode -c tests/pbf/bins-0-0-0.pbf 0 0 0 | grep 'clustered:sum:POP1950' | sed 's/.*"clustered:sum:POP1950": //' | awk '{sum += $$1} END {print sum}'` == 161590
+4
View File
@@ -93,6 +93,7 @@ size_t limit_tile_feature_count_at_maxzoom = 0;
unsigned int drop_denser = 0;
std::map<std::string, serial_val> set_attributes;
unsigned long long preserve_point_density_threshold = 0;
unsigned long long preserve_multiplier_density_threshold = 0;
long long extend_zooms_max = 0;
int retain_points_multiplier = 1;
std::vector<std::string> unidecode_data;
@@ -3109,6 +3110,7 @@ int main(int argc, char **argv) {
{"cluster-distance", required_argument, 0, 'K'},
{"cluster-maxzoom", required_argument, 0, 'k'},
{"preserve-point-density-threshold", required_argument, 0, '~'},
{"preserve-multiplier-density-threshold", required_argument, 0, '~'},
{"Dropping or merging a fraction of features to keep under tile size limits", 0, 0, 0},
{"drop-densest-as-needed", no_argument, &additional[A_DROP_DENSEST_AS_NEEDED], 1},
@@ -3320,6 +3322,8 @@ int main(int argc, char **argv) {
}
} else if (strcmp(opt, "preserve-point-density-threshold") == 0) {
preserve_point_density_threshold = atoll_require(optarg, "Preserve point density threshold");
} else if (strcmp(opt, "preserve-multiplier-density-threshold") == 0) {
preserve_multiplier_density_threshold = atoll_require(optarg, "Preserve multiplier density threshold");
} else if (strcmp(opt, "extend-zooms-if-still-dropping-maximum") == 0) {
extend_zooms_max = atoll_require(optarg, "Maximum number by which to extend zooms");
} else if (strcmp(opt, "retain-points-multiplier") == 0) {
+1
View File
@@ -65,6 +65,7 @@ extern long long extend_zooms_max;
extern int retain_points_multiplier;
extern size_t maximum_string_attribute_length;
extern std::string accumulate_numeric;
extern unsigned long long preserve_multiplier_density_threshold;
struct order_field {
std::string name;
+2
View File
@@ -94,9 +94,11 @@ struct serial_feature {
#define FEATURE_DROPPED -1
#define FEATURE_KEPT 0
#define FEATURE_ADDED_FOR_MULTIPLIER_DENSITY INT_MAX
// <0: dropped
// 0: kept
// >0: sequence number of additional feature kept by retain-points-multiplier
// INT_MAX: additional feature kept by preserve-multiplier-density-threshold
int dropped = FEATURE_DROPPED; // was this feature dropped by rate?
// unsigned long long drop_by; // dot-dropping priority
File diff suppressed because it is too large Load Diff
File diff suppressed because it is too large Load Diff
File diff suppressed because it is too large Load Diff
+31 -7
View File
@@ -1056,7 +1056,8 @@ static bool skip_next_feature(decompressor *geoms, std::atomic<long long> *geomp
// This function is called repeatedly from write_tile() to retrieve the next feature
// from the input stream. If the stream is at an end, it returns a feature with the
// geometry type set to -2.
static serial_feature next_feature(decompressor *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, std::atomic<bool> *within, compressor **geomfile, std::atomic<long long> *geompos, long long start_geompos[], std::atomic<double> *oprogress, double todo, const char *fname, int child_shards, json_object *filter, const char *global_stringpool, long long *pool_off, std::vector<std::vector<std::string>> *layer_unmaps, bool first_time, bool compressed, multiplier_state *multiplier_state, std::shared_ptr<std::string> &tile_stringpool, std::vector<std::string> const &unidecode_data, unsigned long long &previndex) {
static serial_feature next_feature(decompressor *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, std::atomic<bool> *within, compressor **geomfile, std::atomic<long long> *geompos, long long start_geompos[], std::atomic<double> *oprogress, double todo, const char *fname, int child_shards, json_object *filter, const char *global_stringpool, long long *pool_off, std::vector<std::vector<std::string>> *layer_unmaps, bool first_time, bool compressed, multiplier_state *multiplier_state, std::shared_ptr<std::string> &tile_stringpool, std::vector<std::string> const &unidecode_data, unsigned long long &previndex, double droprate) {
double extra_multiplier_zooms = log(retain_points_multiplier) / log(droprate);
while (1) {
serial_feature sf;
long long len;
@@ -1085,6 +1086,18 @@ static serial_feature next_feature(decompressor *geoms, std::atomic<long long> *
sf = deserialize_feature(s, z, tx, ty, initial_x, initial_y);
sf.stringpool = global_stringpool + pool_off[sf.segment];
// with fractional zoom level, so we can target a specific number
// of features to keep with retain-points-multiplier, not just the
// powers of the drop rate
//
// the shift-right is because we lose the bottom two bits of
// point coordinate precision in the calculation in serial.cpp.
//
// the fractional part should actually be scaled in an exponential
// curve instead of linear, but I can't figure out how to make it
// come out right, and this should be close enough.
double feature_minzoom = sf.feature_minzoom - (bit_reverse(sf.index >> 2) / pow(2, 64));
size_t passes = pass + 1;
double progress = floor(((((*geompos_in + *along - alongminus) / (double) todo) + pass) / passes + z) / (maxzoom + 1) * 1000) / 10;
if (progress >= *oprogress + 0.1) {
@@ -1218,12 +1231,16 @@ static serial_feature next_feature(decompressor *geoms, std::atomic<long long> *
sf.dropped = FEATURE_KEPT; // the first feature in each tile is always kept
}
if (z >= sf.feature_minzoom || sf.dropped == FEATURE_KEPT) {
if (z >= feature_minzoom || sf.dropped == FEATURE_KEPT) {
count->second = 0;
sf.dropped = FEATURE_KEPT; // feature is kept
} else if (count->second + 1 < retain_points_multiplier) {
} else if (z + extra_multiplier_zooms >= feature_minzoom && count->second + 1 < retain_points_multiplier) {
count->second++;
sf.dropped = count->second;
} else if (preserve_multiplier_density_threshold > 0 && count->second + 1 < 2 * retain_points_multiplier &&
sf.gap > ((1LL << (32 - z)) / preserve_multiplier_density_threshold) * ((1LL << (32 - z)) / preserve_multiplier_density_threshold)) {
count->second++;
sf.dropped = FEATURE_ADDED_FOR_MULTIPLIER_DENSITY;
} else {
sf.dropped = FEATURE_DROPPED;
}
@@ -1287,6 +1304,7 @@ struct run_prefilter_args {
std::vector<std::string> const *unidecode_data;
bool first_time = false;
bool compressed = false;
double droprate = 1;
};
void *run_prefilter(void *v) {
@@ -1297,7 +1315,7 @@ void *run_prefilter(void *v) {
unsigned long long previndex = 0;
while (1) {
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->start_geompos, rpa->oprogress, rpa->todo, rpa->fname, rpa->child_shards, rpa->filter, rpa->global_stringpool, rpa->pool_off, rpa->layer_unmaps, rpa->first_time, rpa->compressed, &multiplier_state, tile_stringpool, *(rpa->unidecode_data), previndex);
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->start_geompos, rpa->oprogress, rpa->todo, rpa->fname, rpa->child_shards, rpa->filter, rpa->global_stringpool, rpa->pool_off, rpa->layer_unmaps, rpa->first_time, rpa->compressed, &multiplier_state, tile_stringpool, *(rpa->unidecode_data), previndex, rpa->droprate);
if (sf.t < 0) {
break;
}
@@ -1804,6 +1822,7 @@ long long write_tile(decompressor *geoms, std::atomic<long long> *geompos_in, ch
rpa.unidecode_data = &unidecode_data;
rpa.first_time = first_time;
rpa.compressed = compressed_input;
rpa.droprate = arg->droprate;
// this does need to be a real thread, so we can pipe both to and from it
if (pthread_create(&prefilter_writer, NULL, run_prefilter, &rpa) != 0) {
@@ -1835,7 +1854,7 @@ long long write_tile(decompressor *geoms, std::atomic<long long> *geompos_in, ch
ssize_t which_serial_feature = -1;
if (prefilter == NULL) {
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, start_geompos, &oprogress, todo, fname, child_shards, filter, global_stringpool, pool_off, layer_unmaps, first_time, compressed_input, &multiplier_state, tile_stringpool, unidecode_data, next_feature_previndex);
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, start_geompos, &oprogress, todo, fname, child_shards, filter, global_stringpool, pool_off, layer_unmaps, first_time, compressed_input, &multiplier_state, tile_stringpool, unidecode_data, next_feature_previndex, arg->droprate);
} else {
sf = parse_feature(prefilter_jp, z, tx, ty, layermaps, tiling_seg, layer_unmaps, postfilter != NULL);
}
@@ -1883,8 +1902,11 @@ long long write_tile(decompressor *geoms, std::atomic<long long> *geompos_in, ch
} else {
can_stop_early = false;
if (sf.dropped != FEATURE_DROPPED) {
if (sf.dropped != FEATURE_DROPPED && sf.dropped != FEATURE_ADDED_FOR_MULTIPLIER_DENSITY) {
// Does the current multiplier cluster already have too many features?
// (Because we are dropping dynamically, and we have already filled the
// cluster with features that were dynamically dropped from being
// primary features)
// If so, we have to drop this one, even if it would potentially qualify
// as a secondary feature to be exposed by filtering
if (layer.multiplier_cluster_size >= (size_t) retain_points_multiplier) {
@@ -1894,7 +1916,7 @@ long long write_tile(decompressor *geoms, std::atomic<long long> *geompos_in, ch
}
if (sf.dropped == FEATURE_DROPPED || drop_rest) {
multiplier_seq = (multiplier_seq + 1) % retain_points_multiplier;
multiplier_seq = 0;
if (find_feature_to_accumulate_onto(features, sf, which_serial_feature, layer_unmaps, LLONG_MAX, multiplier_seq)) {
preserve_attributes(arg->attribute_accum, sf, features[which_serial_feature]);
@@ -2131,6 +2153,8 @@ long long write_tile(decompressor *geoms, std::atomic<long long> *geompos_in, ch
if (sf.dropped == FEATURE_KEPT) {
layer.multiplier_cluster_size = 1;
lead_features_count++;
} else if (sf.dropped == FEATURE_ADDED_FOR_MULTIPLIER_DENSITY) {
other_multiplier_cluster_features_count++;
} else {
layer.multiplier_cluster_size++;
other_multiplier_cluster_features_count++;