Question

I am fairly new to using Linq which is turning out a bit tricky for me on filtering the best datasource from a c# datatable. I need to perform the following filtering to a datatable. The background behind the data is that it is a set of records which contain independent redundancies from different datasources for the purpose of failsafe. In case one datasource had become corrupted at some point, a secondary or third datasource would become the primary source for that event thread.

Original Data (for example):
signintable:

source  First       Last
d1      John        Smith
d1      John        Smith
d3      John        Smith
d1      Jane        Doe
d2      Jane        Doe
d3      Richard     Miles
d3      Richard     Miles
d1      Richard     Miles

I would like to add two columns to this: a count of unique members of groups by(firstname, lastname and datasource), and a uniqueRecordGroupnumber based on distinct group (firstname, lastname, datasource) but ordered by whichever datasource for that particular first name last name has the most records, to the least.

source  First   Last    Count   UniqueRecordGroup
d1      John    Smith   2       1
d1      John    Smith   2       1
d3      John    Smith   1       2
d1      Jane    Doe     1       1
d2      Jane    Doe     1       2
d3      Richard Miles   2       1
d3      Richard Miles   2       1
d1      Richard Miles   1       2

Then I would finally filter out only the primary (unique record group 1) in order to eliminate the redundancies/ less reliable datasource for that particular record.

source  First   Last    Count   UniqueRecordGroup
d1      John    Smith   2       1
d1      John    Smith   2       1
d1      Jane    Doe     1       1
d3      Richard Miles   2       1
d3      Richard Miles   2       1

How are the above steps accomplished using Linq on a datatable (for example datatable signintable)?

Thank you.

Answer 1

After now completing this application feature, I would not recommend trying to use Linq for this purpose. I found that there were not many resources and forums expressly for more involved Linq queries, whereas with Oracle, or SQL, plenty of forums and posts at least on the parts that were needed. So rather than add the feature in the c#/asp.net layer in Linq script, I added it in the database query layer.

In Oracle, I accomplished the above by 4 steps:

--1.  creating a derived table of unique instances of main data (a): (First, Last) and a count of instances in the main table:

b As
(select First, Last, source, COUNT(*) distinctpersoncount
From a
GROUP BY First, Last, source)

--2.  Rank the fills from greatest to least.

c As
(select b.*, row_number() over(partition by  First, Last
order by distinctpersoncount desc) as distinctpersonrowid
from b


--3.  combine the new distinctperson columns into a joined table with the original data a.

d As
(select a.*, c.distinctpersoncount, c.distinctpersonrowid
from a
left join c
on
.... (join conditions here)

--4.  select all the data in d, filtering by distinctpersonrowid = 1 (the best source).