powered by CADENAS

Manual

Manual

1.7.5.9.2. Block [settings] Key PartialSearchCatalogs - Partial Search
Precondition

The partial search is active by default.

However, the fingerprints for the partial search for a catalog are only created if configured. To create the fingerprints for the catalog "mycatalog", for example, set the following key in geomsearch.cfg:

PartialSearchCatalogs=cat/mycatalog

If you want to set more than one catalog, separate the entries by comma.

This setting applies to every way fingerprints can be created.

Finally create the geometrical index for the whole catalog.

Setting Accuracy

The accuracy is configurable:

$CADENAS_SETUP/geomsearch.cfg

[PartialSearchAccuracies]
Accurate=0.02,Accurate
Intermediate=0.07,Intermediate
Rough=0.15,Rough

First value is similarity (0.10 = 10%) and second value is the visible name in application (translatable in .def).

Background information

The generation of fingerprints for the partial search is significantly more time consuming than the generation of normal fingerprints. Furthermore the fingerprints need more disk space. That's why they are not generated by default for all catalogs.

The index is located in $CADENAS_DATA/index/cat/cat_<catalogname>/partial. It contains the following files:

  • geomsearch.fdb: Fingerprints and attributes for all features

  • geomsearch.ofm: Mapping from project path to internal ID

  • geomsearch.edx: Feature fast search index

  • geomsearch.cidb: Additional information only for indexing

After an update three more files are created:

  • geomsearch.uop: Update information

  • geomsearch.ufdb: Fingerprints and attributes for all updates

  • geomsearch.ucidb: Additional information for all updates

It is built up similar to the normal GeoSearch. Instead of the linear index it uses a fast search index (.edx).

The size of the caches can be altered in the blocks CACHEV2_PARTIAL_SEARCH_32 resp. CACHEV2_PARTIAL_SEARCH_64. Rule of thumb: SampleLineCacheSize has the lowest priority. Adding a valid filename under LogFileName will write cache information after each search. These can be used to distribute the memory optimal.