tandon updated this revision to Diff 5222.
tandon marked 7 inline comments as done.
tandon added a comment.


  Compared to last time, results were quite according to expectations. Now  all 
the streets, which were previously rendered as broken now appear continuous. 
Also there good savings in terms of file size.( I have presented a few memory 
related statistics below)
  
  **Fixes Done**
  
  - Using OSM node ID's instead of lon-lat coordinates as key values for the 
multi-hash-map
  - Added multi-hash-map to handle intersections, hence now corresponding to a 
node-id there can be 2 or more WayChunks
  - Using visual categories for concatenating/(differentiating among WayChunks)
  - Added reverse() method in GeoDataLineString
  - Refactored code to more modular units.
  - All the variable definitions are now near to where they are used.
  - Fix bugs related to improper reversal of WayChunks, concatenation of 
WayChunks having the same first and last IDs, proper insertion/deletion of 
key-value pairs in multi-hash-map during the execution of the tool.
  
  **Fixes Left**
  
  - Proper memory management.
  - Performing timing related tests.
  - Running the output file through the tile creation tool chain.
  
  **Doubts**
  There is a case in which a Way under consideration is inserted between 2 way 
chunks.
  Example: Suppose Way under consideration is <4,5> and there exists WayChunks 
<1,2,3,4>; <5,6,7> then the program will construct a WayChunk <1,2,3,4,5,6,7>
  Now consider the following scenario:
  Way under consideration is <1,8,9,10,7> and there exists a WayChunk 
<1,2,3,4,5,6,7> then the program generates  nonsensical WayChunks. For now I am 
detecting such scenarios and directly adding the way under consideration to the 
output file hence removing this way from any future considerations.
  However due to this abrupt removal of such wayswe are getting a small 
variance in the final number of ways reduced ,every time we run the program on 
the same input file. Though this variance is not affecting rendering , this 
thing does result in a smaller total number of ways reduced.

REPOSITORY
  rMARBLE Marble

CHANGES SINCE LAST UPDATE
  https://phabricator.kde.org/D2128?vs=5054&id=5222

REVISION DETAIL
  https://phabricator.kde.org/D2128

AFFECTED FILES
  src/lib/marble/geodata/data/GeoDataLineString.cpp
  src/lib/marble/geodata/data/GeoDataLineString.h
  tools/osm-simplify/BaseFilter.cpp
  tools/osm-simplify/BaseFilter.h
  tools/osm-simplify/CMakeLists.txt
  tools/osm-simplify/TagsFilter.cpp
  tools/osm-simplify/TagsFilter.h
  tools/osm-simplify/WayChunk.cpp
  tools/osm-simplify/WayChunk.h
  tools/osm-simplify/WayConcatenator.cpp
  tools/osm-simplify/WayConcatenator.h
  tools/osm-simplify/main.cpp

EMAIL PREFERENCES
  https://phabricator.kde.org/settings/panel/emailpreferences/

To: tandon, rahn, nienhueser
Cc: rahn, #marble, marble-devel
_______________________________________________
Marble-devel mailing list
[email protected]
https://mail.kde.org/mailman/listinfo/marble-devel

Reply via email to