[Kmymoney-devel] KMyMoney - Matching Issue with CSV Importer Plugin

[allan]

On 14/08/10 16:53, allan wrote:
> On 14/08/10 12:04, allan wrote:
>   
>> I'm nearly there with the CSV Importer Plugin, but have found an issue
>> with one of my test files.  It was sent to me a while ago by a lister
>> who needed to import a csv file from a Spanish bank, and was provided
>> mainly as an illustration of the format, but has probably been 'edited',
>> and perhaps is not exactly a likely situation, but shows a
>> not-impossible one.
>>
>> The issue concerns item matching.  The file contains several duplicated
>> transaction occurring on the same date, and can be boiled down to this:-
>>
>> 13/08/2010,payee1,10.00
>> 13/08/2010,payee2,10.00
>> 13/08/2010,payee3,10.00
>>
>> The csvimporter can output either via the plugin facility directly into
>> KMM, or alternatively produce a QIF file for import.  With this test
>> file I get different results (bearing in mind that both have the same
>> source.)
>>
>> Apart from the fact that they get assigned an unconnected payee (an
>> issue I've reported previously), what happens with the csv plugin import
>> is that I get all three imported, but with a match occurring on two of
>> them.  In Ledger view, I see 'KMyMoney has matched a downloaded
>> transaction with a manually entered one... .  The third identical one is
>> shown separately.  No duplicate is flagged.
>>
>> When imported via QIF, all three items show up as separate.
>>
>> With a real-life import of 29 items, all that should match, do match. 
>> There is though an occurrence of two items that were imported being
>> matched against each other when they should not have been.  These were
>> on the same day, and for the same amount, but with a differing
>> payee/detail/description field.
>>
>> Why do two imported transactions get matched, as though one were
>> manually entered and the other imported?
>>
>> Thanks
>>
>> Allan
>>     
> Solved, more or less.  I'd spent a few hours hunting around, but a
> further half hour gave me an answer.  No bank ID.  For the moment, I've
> just used a rand for that and all is now working well, pending further
> research.
>
> Allan
>   


Just wanting confirmation that it's OK to use qrand to produce the bank
ID.  Initially, I used this as a quick attempt to see if I was on the
right lines with my problem.  Apart from fixing the immediate issue, it
otherwise seems OK, too.  Further, I see in mymoneyqifreader.cpp that,
apart from using the hash/qmap, where a '#' is encountered, a value from
the qif file may be used, I think, which suggests that it's not
necessary to use qmap to produce it.

I see no explanation for why that particular method id used, but don't
want to screw things up further along the line.

Allan