Last week i was speaking about a new solution for the merge. ( As reminder : I work on node splitting where there is too much data into one one ).
The splitting work perfectly and i good to go.
For the merge, i've tried the second solution i was speaking of last week, which was let the node expire.
But here comes another problem, when node expire and if they are at the end of the trie, then all above node will expire too ... and then data who was at the end of the trie will be unreachable.
Then i need to comes up with another solution ...
For the moment, since i go none, i will juste end with the splitting by including a "move" of the data instead of the reinsert which mean i need to keep track of the timer of the data and reinsert them with the same time_point where they where insert.
I also will write an article for the blog of Savoir Faire Linux ( https://blog.savoirfairelinux.com/en/ ).
The article should be done for Friday :D
After it i will try other solution for the merge of the data and will then start working of the multidimensional data indexation.
Anyway, see you next week, with hopefully some good news for this merge who don't want to merge :D
Subject: FW: GSoC 2016 Week 5 - Indexation and data Persistance over OpenDht
Date: Thu, 30 Jun 2016 18:41:55 +0200
Last week i was speaking about my new solution for the data persistance, now there is an implementation of that.
But here comes another problem :
If there is a lot of same value, then the split be done until the end ( that ok ), but it will not put the value by the end of the triee since there is not enough value.
That will happend is the following :
There problem as you may guest is that the data does a kind of "ping-pong" between last nodes and we dont want that since that will give a lot of useless data transfer.
This week i work on another method for the merge which is basically let the node expire.
then, when we put another value, we will node update the node and then after some time it will expire ... and so the data will be put into the parent.
And Voila, merge is done.
Hope that, that will work.
Anyway, see you next week, next week i'll probably start the indexation of data with multi key :D