Null values in F2 of optimised network

[cjacobscrisioni]

/NetworkSetup/Per_Continent/Australia_Oceania/NetwerkSpec/CreateMoreEfficientNetwork/OptimiseNetwork/Iter_2/IntermediateLinkSet/F2 throws 3 null values (after the commit on 14/12, 22:07, SHA 5edc8b510ba51e832c5b5aa6edad2d6220bbd552)

The issue comes into play in the second iteration because a number of JunctionFreeSection ids are not represented in ConnectorLinks/JunctionFreeSection_rel .

Solutions tried:

• Moved entire network and nodeset to Webmercator (hypothesis: to avoid potential issues with WGS84 not being discerning enough geographically)
• Set all OneDirections to False (hypothesis: weird closed-off sections due to haphazard onedirection codes) Both solutions did not work.

The issue is currently circumvened (after the commit on 15/12. 08:26, SHA 77a05e511e002a6422e0aafeb5751005b91d2509) but the cause is unknown. This merits a deeper look to understand the cause properly.

[jipclaassens]

Indeed now there are no issues

[jipclaassens]

[jipclaassens]

https://github.com/ObjectVision/AccessibilityModelling/wiki/Network-optimisation-algorithm

[jipclaassens]

de min(id(connector_link), connector_link/jfs_rel) gaf een undefned getal weer +4.2miljard, of 0 in het geval van max(). Dat is niet goed. Die zijn iig nu vervangen met first en last.

[jipclaassens]

Nu denk ie een jfs te zijn, maar geen connector links te hebben.

[jipclaassens]

/NetworkSetup/Per_Continent/Australia_Oceania/NetwerkSpec/CreateMoreEfficientNetwork/OptimiseNetwork/Iter_1/JunctionFreeSection/First_ConnectorLink_rel

In JFS record id 3018 has null for first_connector link and many other attributes.

first_connector_rel := first(id( ConnectorLink ), ConnectorLink/JunctionFreeSection_rel ) But ConnectorLink/JunctionFreeSection_rel has no id 3018... ?!

How come a record in JFS is created for something that doesn't exist??!

[jipclaassens]

It might be due to unconnected parts in the network. Now added a connectiveness script after creating the initial network. Based on WebMercator. En still a lot of unconnected parts

[cjacobscrisioni]

The network is expected to consist of multiple unconnected parts by design - else we would drop all roads in eg Fiji. The rule is to accept small subnetworks if they happen to be the only one connected to a sep_unit (ie an island)

[jipclaassens]

Yes, i know. But still with your script there are subnetworks.

Still investigating, tbc

[jipclaassens]

Here, only connected links are selected (based on Chris' script, which takes islands/countries into consideration). /NetworkSetup/Per_Continent/Australia_Oceania/NetwerkSpec/CreateInitialWorkingNetwork/Roads_Connected

Next, the connect-operator is applied.

Then, I calculated the connectedness again on the result. And now there are unconnected links again...

[jipclaassens]

We have found it! Cleaning results in a cycle. And a cycleis again a JunctionFreeSection and thus can be cleaned again. But in this case we did not omit this in time.

[jipclaassens]

Fixed!

[cjacobscrisioni]

ahhhh whahoo congrats !!!!