Interesting. I'm not sure if this is a bug in Nominatim or a misunderstanding of OSM definitions. Let's see: - Dilermando de Aguiar is a Brazilian municipality with 3308 inhabitants, mapped as an administrative boundary with admin_level=8 with a place=village as its admin_centre node - Sede is a municipal district, mapped as an administrative boundary with admin_level=9 sharing the same admin_centre node from its parent municipality (because it's the main district) - São José da Porteirinha, currently mapped as as place=hamlet node, is an urban agglomeration that is spatially disconnected from the main urban agglomeration of its municipality, with no special political or administrative status
For each enclosing area, Nominatim computes a rank, using distinct rules for each country [1], and forms the address line by concatenating comma-separated names in descending rank order. In this particular case, it computed [2] rank 16 for São José da Porteirinha and for Dilermando de Aguiar, and rank 18 for Sede. Thus, it concluded that Sede (rank 18) is contained in any of the other two (rank 16), of which only one was chosen. The documentation of Nominatim [1] suggests that it assumes that a hamlet will never exist within a village, since by definition in OSM only certain types of places may be parts of other settlements [3]. It could have worked if Sede was place=town/city (possibly incorrect or more correct [4][5]) or if São José da Porteirinha was place=neighbourhood, which depends on considering the latter distinct or not from other settlements. From the text in the wiki, I understand that any given small piece of land can be considered part of an urban settlement, part of a rural settlement, or unsettled land. In Brazil, all municipalities, all municipal districts and the neighbourhoods with boundaries defined legally can include any of these types of areas and therefore do not correspond perfectly to the concept of settlement. If municipalities are always considered fully settled, any place inside a municipality must be of a type that may be part of another settlement (such as a suburb or neighbourhood) and not a type of distinct settlement (such as a hamlet or village), no matter if there's unsettled land between remote places and the nucleus of the municipality. If municipalities are not considered fully settled, it could be argued that their boundary should not cover unsettled areas (probably undesirable). If none of the interpretations are correct, Nominatim's ranking system must have several bugs. On a side note, Nominatim's default ranking [1] seems to suggest that towns could be part of cities, though the wiki [3] does not explicitly suggest this. [1] https://wiki.openstreetmap.org/wiki/Nominatim/Development_overview#Indexing.2Faddress_calculation [2] https://nominatim.openstreetmap.org/details.php?place_id=4150595 [3] https://wiki.openstreetmap.org/wiki/Key:place#Populated_settlements.2C_urban [4] https://en.wikipedia.org/wiki/Subprefecture [5] (pt) https://forum.openstreetmap.org/viewtopic.php?pid=702045#p702045 On Mon, Jun 4, 2018 at 5:06 PM, santamariense <[email protected]> wrote: > Hello everyone, > > In Brazil the tag place=hamlet is being used for smallest places, however, > according to Wikipedia, hamlet may be an unincorporated place. And, > I've always observed that place=hamlet isn't in the correct hierarchy > when it's shown in Nominatin. > > Example: São José da Porteirinha ( > https://www.openstreetmap.org/node/416766430 ) is shown replacing the > name of the municipality (admin_level=8) in Nominatin: "Sede, São José > da Porteirinha, Microrregião de Santa Maria, Mesorregião > Centro-Ocidental Rio-Grandense, Rio Grande do Sul, Região Sul, Brasil" > ( https://www.openstreetmap.org/search?query=São%20José%20da%20Porteirinha > ), where, the correct hierarchy should be "São José da Porteirinha, > Sede, Dilermando de Aguiar, Microrregião de Santa Maria, Mesorregião > Centro-Ocidental Rio-Grandense, Rio Grande do Sul, Região Sul, Brasil" > > Then, I believe that for Nominatin, all place=hamlet is being taken as > unincorporated place. There's no place in Brazil that is > unincorporated. By the correct hierarchy, place=hamlet in Brazil > should be hierarchly got after level=10. > > That's so weird as saying that Nominatin should show a hierarchy like > State, City, Country instead of City, State, Country. > > The question is: Is Nominatin wrong or are we mapping a thing that > there not be in Brazil and a best tag should be applied? Or maybe a > new tag like unincorporated=yes/no should be created to complement the > type of hamlet? > > []'s > > user santamariense > > _______________________________________________ > Tagging mailing list > [email protected] > https://lists.openstreetmap.org/listinfo/tagging -- Fernando Trebien _______________________________________________ Tagging mailing list [email protected] https://lists.openstreetmap.org/listinfo/tagging
