hand-coded transactions are usually BAD transactions

[Kevin HaleBoyes]

Josh Sled wrote:
> On Sat, 2004-10-30 at 07:36, Neil Williams wrote:
> 
> 
>>By hand? There's no telling the chaos that will result in the QofBook when you 
>>arbitrarily insert items. Did you check the GUID? did you create them out of 
>>thin air? Did you make any effort to match the GUID of the parent against the 
>>GUID of the child?
> 
> 
> It's not magic, let alone chaotic.
> 
> If you randomly create a GUID [in the expected format, of course] from a
> reasonable random-number source, there's a _really_ small chance you'll
> create an existing one.  If you want to be paranoid, you could check it
> against all known GUIDs in the same dataset.

I just replied to Neil's concerns, and there I pointed out that I did
use the GUID from existing transactions that are similar to the ones
I tried to create.

>>>First, is it possible to even do this?
>>
>>Not by hand, at least, not without a very high probability of errors.
> 
> 
> Sure it is, with a pretty low probability of error...  it's not
> recommended, but it is possible:
> 
> 1/ Look in the datafile and understand the <gnc:transaction> blocks.
> 
> 2/ Create another one with the correct contraints:
>   * The transaction id [<trn:id>] GUID must be unique.
>   * The Split [<split:id>] GUID should be, too.
>   * The <split:account> guids should reference the appropriate account.
>   * The value/quantity should be valid.

Ok, I think that is what I missed.  The <trn:id> or <split:id> were
not unique in my generated transactions.

> But you're probably going to find it easier and safer to just create QIF
> files and use the existing import mechanism, as Derek suggested.

Derek also suggested that.  It is probably the more sensible way to
approach this.


Thanks for your time,
K.