[pgrouting-dev] Time dependent data and input format
Jay Mahadeokar
jai.mahadeokar at gmail.com
Fri Jul 29 11:49:04 EDT 2011
Hi Steve,
On Fri, Jul 29, 2011 at 6:26 PM, Stephen Woodbridge <woodbri at swoodbridge.com
> wrote:
> Jay,
>
> This make sense. What I'm trying to understand and get to is how do we make
> this easy to setup and use. It seems like we are missing something
> somewhere. For example, the cyclical range of a single entry can be
> different from other entries, think week days versus weekends. It should be
> possible to have some cyclic entries and some not cyclic. Based on this it
> seems that the cyclic flag needs to be set on each record along with the
> period of the cycle. Would you agree?
>
>
The main motivation behind all this brain storming is to keep the core
algorithm and the postgres c function implementation, independent of the
nature of data, right? So, we need to somehow abstract the data dependent
properties from the core functionality.
That was the reason why we came up with the plsql wrapper functions idea.
For optimising again, we need to communicate the cyclic properties to core
algorithm, but the interface should be independent of nature of data.
> If this is the case, then a quick scan of the data would tell us if there
> are cyclic entries, in which case a simple function to scan the data needed
> and see if any entry is cyclic could return that flag. But if every entry is
> marked as cyclic or not then you might not need that flag at all.
>
What I feel is: if we scan the data and keep the logic of determining the
cyclic nature within the postgres C function, then the above aim will
somehow get lost and it will again become a data format specific function.
So, I think it will be better if the application developers deal with the
data themselves so that they can achieve the optimisation. Example - for
weekday, weekend, the plsql function will provide specific values for each
day and mark data as non cyclic if there is transition from weekday to
weekend and vice versa. (A journey more than few days is highly unlikely, so
anyways it wont repeat the data too much, though it will not be optimal)
If the journey is within weekday or within weekend, then we can mark data as
cyclic, and just retrieve 24 hour data.
This is specific to example data you specified. I believe such
improvisations will be applicable for most data formats. Thoughts?
Again, this is not the most optimal implementation, but still it keeps the
core implementation independent of data format. Any other ideas?
>
> May be I'm missing something because I have not had a full cup of tea yet
> :)
>
> Thoughts,
> -Steve
>
>
> On 7/29/2011 8:39 AM, Jay Mahadeokar wrote:
>
>>
>>
>> On Fri, Jul 29, 2011 at 5:50 PM, Stephen Woodbridge
>> <woodbri at swoodbridge.com <mailto:woodbri at swoodbridge.**com<woodbri at swoodbridge.com>>>
>> wrote:
>>
>> On 7/29/2011 5:49 AM, Jay Mahadeokar wrote:
>>
>> Hi all,
>>
>> I was looking at ways to optimise the tdsp implementation.
>>
>> Currently, if the time-interval is suppose 24, (24 hours) and
>> the travel
>> time expected is around 60 hours, then the tdsp-wrapper query
>> retrieves
>> total 60 intervals and the data is repeated for the 24 hour
>> intervals.
>> (I hope what i am trying to say is clear).
>>
>> This is unnecessary space wastage. So, I edited the weight_map
>> class to
>> accommodate a variables is_cyclic and the cycle_interval, which
>> will
>> keep track of the cyclic nature of the data. So, actually only
>> the 24
>> hour data will be fed to weight map and then it will be reused
>> in cycles.
>>
>> Because of the good design, I guess this has scaled up really
>> nice and
>> there is no need to modify actual core tdsp. Just the weight_map's
>> get_travel_time() function and the wrapper plsql function needed
>> to be
>> altered.
>>
>>
>> This makes perfect sense and sounds like a great optimization.
>>
>>
>> Now, in the main query, I am thinking of adding one more boolean
>> variable - is_cyclic which will be enabled if the data is cyclic in
>> nature. Should we keep this? Or we should assume that the data
>> will be
>> always cyclic.
>> Any other views on the same are welcome.
>>
>>
>> I guess my questions are:
>> How does this get set?
>> What are the advantages of setting it?
>> Is there a way to dynamically check if it is cyclic at the start and
>> avoid having the app builder set it up?
>>
>>
>> Advantages of setting it:
>> If the data is not cyclic, then the whole data will be expected to
>> reside in database itself. So, if the upper bound of the time is reached
>> then by default, we can assume that the travel_time corresponding to the
>> last existing time window should be used for such cases.
>>
>> If data is cyclic, then we need to start all over again. These are the
>> cases, what I could think of.
>>
>> How does this get set:
>> The tdsp query must have an additional parameter - is_cyclic which
>> should be used to set the is_cyclic flag of the weight_map.
>>
>> Now, if data is cyclic, then the plsql function that retrieves the data
>> should be written in such a way that if interval is suppose 0 - 24, and
>> start time is 21, then 21->0 and 20->23. After that the cyclic flag
>> will indicate us to loop again. The plsql function will be generally
>> written by app developer and he should take care of this. (I have
>> already implemented this in my code.)
>>
>> If data is cyclic, he should pass the flag appropriately. If is_cyclic
>> is false, we can assume that if intervals are exhausted, we will keep
>> using time corresponding to last interval.
>>
>> Is this reasonable?
>>
>> I suppose it is not a big deal for the app builder to set this flag
>> since he will know if the data loaded is cyclic in nature. I assume
>> you are only looking for some indication if there are cyclic entries
>> in the data and not if this specific query has a cyclic nature which
>> would be impossible to know without analyzing the start time and max
>> time window, in which case we should figure that out automatically
>> because the user making the request is not likely to know anything
>> more than start, end and start time or end time.
>>
>> -Steve
>>
>> --
>> Regards,
>> -Jay Mahadeokar
>>
>>
>>
>> ______________________________**_________________
>> pgrouting-dev mailing list
>> pgrouting-dev at lists.osgeo.org
>> http://lists.osgeo.org/**mailman/listinfo/pgrouting-dev<http://lists.osgeo.org/mailman/listinfo/pgrouting-dev>
>>
>
> ______________________________**_________________
> pgrouting-dev mailing list
> pgrouting-dev at lists.osgeo.org
> http://lists.osgeo.org/**mailman/listinfo/pgrouting-dev<http://lists.osgeo.org/mailman/listinfo/pgrouting-dev>
>
--
Regards,
-Jay Mahadeokar
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.osgeo.org/pipermail/pgrouting-dev/attachments/20110729/a92a0507/attachment.html
More information about the pgrouting-dev
mailing list