arjunmehta/node-georedis

georedis with own storage

Opened this issue · 4 comments

measwel commented

Good day,

I would like to implement geohashing based on your module, but using own storage ( not redis ).
Currently I store locations as int geohash (bit depth 52). For one location this is: 3672837726667608

By using the module I can calculate ranges based on location and radius. The result is:

[ 3672437556248576, 3672506275725312 ],
[ 3672574995202048, 3672918592585728 ],
[ 3672987312062464, 3673056031539200 ],
[ 3673193470492672, 3673330909446144 ]

My question:

  1. How should the query to the database look like? How should I compare the stored locations to the calculated range values? Should I just have 4 conditions testing if the geohash falls between the range boundaries?
  2. Why are there 4 ranges and not 8?
  3. Is searching within 4 ranges actually searching within a rectangle and not within a circle? ( the ranges have been calculated based on a radius.

Thank you.

arjunmehta commented

@measwel You can look at my answer here at gis.stackechange.

You will need to find a way to query a range of numbers. As long as you can do that, you should be able to implement this using your own form of storage.

Hope this helps!

measwel commented

Hallo Arjun,

Thank you. I managed to implement integer geohash search using my own storage :)
However, I get results which I cannot explain.

  1. The smaller the search radius, the longer the query takes. I am querying 100.000 random points. With 100km radius, the query takes 5 seconds to execute. However, with 200km radius, execution time drops to 80 ms. Is this normal?

  2. When I calculate the distance between the central point of the search and the results, I get values bigger than the search radius. Searching with 200km radius returns points which are almost 400km away. Is this normal?

arjunmehta commented

That seems very wrong. The query shouldn't take much longer with different radii with a properly implemented ordered set. You need to be able to query your set with specific ranges.

measwel commented

I simply run 4 queries ( one for each range I get ) and aggregate the results.

I have a feeling that the errors I see might have to do with bigint implementation. I am considering switching to integer (32 bit) storage of integer hashes. How much precision can I expect in meters for 32 and for 64 bit integer storage? Can you provide me with an estimate of the maximum error I can expect?

Thank you!

The getQueryRangesFromRadius function for:
lat lon radius accurate
INPUT: 52.582014799999996 13.4913122 10000 true

getQueryRangesFromBitDepth(lat, lon, rangeBitDepth, 52);

Produces:
RANGES: [ [ 3672810144661504, 3672811218403328 ],
[ 3672813365886976, 3672815513370624 ],
[ 3672833766981632, 3672834840723456 ],
[ 3672835914465280, 3672841283174400 ] ]

Could you please check if I am getting the correct integer values here?