Following up... I do not think there is a simple solution for this.
I have table A and B and a joining table C, with adjoining member
collections in both A and B having many C instances held there.
I load in the domain properly and detach, I have the instances and then let
the ISession etc fall out of scope and dispose.
I am removing an existing C relationship there detatched. I see the correct
bidirectional logic happening, at least as far as the domain is concerned.
Then I feed these instances back into a new session and am trying to run
either Save or SaveOrUpdate and getting this error.
NHibernate.PropertyValueException: 'not-null property references a null or
transient value WhalleyBotEnhanced.Console.Spike.Models.MyRoster.Team'.
That I know of, there are no database requests happening during this time,
things are entirely ORM bound at that moment.
Let's talk practical model terms. For brevity, obviously there are backing
fields and such invovled, but for sake of conversation, this is the desired
shape.
class League
{
IList<Team> Teams { get; set; }
IList<Player> Players { get; set; }
}
class Team
{
League League { get; set; }
IList<Roster> Rosters { get; set; }
}
class Player
{
League League { get; set; }
IList<Roster> Rosters { get; set; }
}
class Roster
{
bool IsStarter { get; set; }
Rank Rank { get; set; }
Team Team { get; set; }
Player Player { get; set; }
}
Obviously there is business logic there, observable collections wrapping
the IList<T> indirectly, allowing NH to engage with the mapped properties.
In my detached sequence I am doing exactly this, which would not be
atypical of a given scenario.
if (leagues.FirstOrDefault() is League league
&& league.Players.FirstOrDefault() is Player player
&& player.Rosters.FirstOrDefault() is Roster roster)
{
var team = roster.Team;
// Remove one player from the team, observable collection responds with
the player-team side.
team.ObservableRosters.Remove(roster);
// Generate a new player and add to leage then team roster.
player = new() { };
league.ObservablePlayers.Add(player);
roster = new() { Player = player, IsStarter = true };
team.ObservableRosters.Add(roster);
}
So there are some mixed persistent and transient instances going on here,
dirties a collection in the process, etc.
I save the objects at the league level.
Do I need to re-query and compare the tree side by side? Then at least
start by deleting instances which would be orphaned or should be deleted,
i.e. the Roster instance. I did something similar with EF actually.
Obviously what I am doing with this experiment is not working, so I need to
do something else different. What I have detached and feeding the UI and
asynchronous processes, is in one state. And will always need to
potentially be reconciled with the database state in some way.
Any clues around this sort of behavior would be grand.
I did scan over the manupulating data chapter, particularly on detached
data, but I'm not sure it was especially helpful. I tried what it suggested
and got these errors.
https://nhibernate.info/doc/nhibernate-reference/manipulatingdata.html
>From other forums, one thought is to restructure Team and Roster, and since
Players are not technically required to join a team, having a team null
instance after a sort to help with that business logic.
Honestly right now that thought may be growing on my regardless, but might
like to maintain the Roster table, as there are other bits of details that
only make sense for that, not for player per se; i.e. Rank, IsStarter, for
instance.
Best.
Michael W. Powell
On Saturday, June 7, 2025 at 12:16:40 AM UTC-4 Michael W Powell wrote:
> I should also mention to clarify further, I have the use case where the
> instances I load are effectively all detached from any session. I do this
> intentionally as it feeds my WPF application. Then when I am ready to save
> after some changes have been made to the instances and constituent
> collections, I would reattach all to a new session. There are moments when
> that happens when the composition of collections would then be different.
> So there needs to be some reconciling between the two. I had to take some
> steps in countermeasure with EF apparently; not sure if FNH is any more or
> less transparent under those conditions.
>
> Thanks!
>
> On Friday, June 6, 2025 at 11:29:51 PM UTC-4 Michael W Powell wrote:
>
>> Hello,
>>
>> Now that I feel I have sorted most of the major concerns and but for a
>> couple of lingering desires for reference conventons and what not...
>>
>> Now to one burning question, how to handle synchronizing with persistent
>> collections. What do I mean by that.
>>
>> I am pretty confident that new instances would simply be inserted and
>> existing instances be updated, as per usual.
>>
>> But what happens with non-existent instances. That have been removed from
>> the loaded and/or newly attached instance, and that still exist in the
>> persistent store?
>>
>> The issue I was findingwith EF, for instance, these instances required
>> manual intervention, or I had not discovered the EF policy making that
>> issue transparently resolved at the ORM level.
>>
>> Does it make sense? Much appreciated any insights.
>>
>> Best, thank you,
>>
>> Michael W. Powell
>>
>
--
You received this message because you are subscribed to the Google Groups
"Fluent NHibernate" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To view this discussion visit
https://groups.google.com/d/msgid/fluent-nhibernate/a41cef6c-2f15-49dd-b2bd-9607b1694136n%40googlegroups.com.
