[GNC] import map editor

[John Ralls]

> On May 12, 2018, at 9:46 AM, Eric Theise <erictheise at gmail.com> wrote:
> 
> Thank for answering my questions, Geert.
> 
> On Sat, May 12, 2018 at 6:21 AM Geert Janssens <geert.gnucash at kobaltwit.be>
> wrote:
> 
>> Do let us know if the import behaves better after you have removed the bad
>> map
>> entries.
>> 
> 
> I only import into each account on a monthly basis so it'll take some time
> to really answer this. For the account I reconciled yesterday I deleted
> many dozens of entries; there were only two that looked correct. As I noted
> in the ticket, there were many cases where account names appeared in the
> Match String which, sans documentation, seems very suspicious to me.
> 
> Many of the new entries for that account look promising but I wish it
> wouldn't create entries for day names. I can see use cases for that feature
> but in my case, day names are noise. There are many one & two character &
> digit Match Strings which will also be noise.
> 
> I wonder if the Non-Bayesian is better for me and if there are paths to
> creating rules. I do wonder if Bayesian and Non-Bayesian can be in use at
> the same time.
> 
> Oops, as I go to look the Preferences window locks up. Time for a different
> ticket.
> 

Bayesian matching tokenizes the description on whitespace. Each time you assign a transaction whose description contains a particular token the count on the corresponding token-account pair is incremented (and the token-account pair is created if it doesn’t already exist). There’s no filtering applied to the tokens; that might be a useful enhancement.

Matching is accomplished by finding all of the token-account pairs matching the tokens in the transaction’s description and then summing the counts by account. The account with the highest sum is assigned as the match. 

Regards,
John Ralls