Query Builder Error when exporting solution in Dynamics 365

Solutions in Dynamics 365 are very useful and there is no doubt about that. One of the latest features is possibility to add single fields, views or forms to the solution, not the whole entity with all its dependencies. That allows us to better control what is transferred and deployed on out staging and production environments. Few times I came across the problem, which is caused clearly by this granularity of solutions:

error

Query Builder Error

The specified field does not exist in Microsoft Dynamics 365

Fortunately this error was in on-premise environment, so I was able to check the tracing logs and the problem has become clear:

“attribute with id=’86e616c0-64fc-413c-9957-470272c3f198′ does not exist”

So, apparently, someone has removed some attribute from the system, but somehow it was orphaned in our solution and is causing the export to fail (no doubt – we are trying to export the solution, so system is getting all the metadata, for all the Solution Components and cannot find one of the component). This is very unfortunate, because we cannot see this missing object in the solution view, it’s only in database in SolutionComponentBase table.¬† In on-premise environment, it’s not a problem – we simply check the id in traces and then go to our database and do:

SELECT FROM SolutionComponentBase¬† where ObjectId = ‘guid of missing component’

This will get us the failing components from all our solutions. We can remove them on DB level (sic!) but if you are afraid that this is not supported (which it is, but really cannot break anything) you can do it using SDK.

But what if we have online instance without access to the database? We can request Microsoft Support for the traces and then we will know which attribute is failing. But if we are in a hurry, we would have to loop through all Solution Components is our solution, check if it exists in the system and remove it from the solution. I believe that in 99% of cases, the missing component would be an attribute, so here is the code that would help you to get rid of this error:

It simply iterates through all attributes in the solution and tries to retrieve metadata for each one of them. If it fails, it simply removes this component from the solution. That should be enough for most of the cases, but if there will be any other component types missing, you will have to modify the code to check for those missing components (which I described in one of my previous posts). I have prepared a very simple program, that iterates through all of the solution components and tries to retrieve every single type – and if it fails it just removes it from the solution. Feel free to use it for any means necessary:

https://github.com/pawelgradecki/SolutionExportFixer

Disable columns in editable grid

Editable grids are a long awaited feature of Dynamics 365. Without any fancy coding, they allow to provide users a nice and easy to use interface, for updating multiple records in an Excel-like experience. We can easily create a subgrid on the form, configure it to use an Editable grid as control and we are good to go, we can edit all the grid columns.

Przechwytywanie

But what if we would like to edit only some of the columns of the grid, not all of them? This is not configurable, either we have an editable grid with all the columns editable, or we don’t have editable grid at all. The only way of handling this with configuration is enabling Field Security Profile – if user will not have privilege to update the field, it will appear as disabled for him. Fortunately there are quite a few client SDK features, that will allow us to achieve our goal with a little bit of JavaScript. First, we should register a function for OnRecordSelect event on our grid:

Przechwytywanie

And my gridRowSelected function looks like that:

This function basically gets all attributes for the selected row and for specified attribute (in my case it was “odx_monthid”) disables the grid cell. Effects are the following:

Przechwytywanie

Great, so only Month column is currently not editable, but we can still edit Amount column. The only drawback of this solution is that this “lock” on the field is shown only when you click on the row (because we registered the function to be fired on OnRowSelect event). Couldn’t it be fired for OnLoad of the form? Unfortunately, according to MSDN documentation:

we cannot run this, when form loads, because the internals of the grid are not yet loaded. Still it is quite an acceptable workaround. Maybe in future releases, editable grid will receive some more features like configuration of disabled columns.

Using early bounds in plugin for multiple entities

Recently I was caught up in the discussion about early bound entities and late bound entities. One of the argument of the “late bound supporters” was the fact, that using late-bound allows¬†reusing the same plugin for multiple entities (so for example create a plugin that runs for all activities or a common plugin for leads/opportunities/accounts). This argument is not valid, I will show you how to write a clean code, that uses modern approach for coding and is perfectly testable.

Let’s start with the late bound version of the plugin to give you an idea what are we trying to achieve here:

The plugin simply adds values of “new_netamount” and “new_margin” fields on the entity and sets field “new_totalsum” to a resulting value. Of course this can be simply achieved by using calculated field, but I just want to show the basic idea here, real world scenario is usually much more complicated.

If we register this field on pre-update of any entity containing field “new_totalsum”, “new_netamount” and “new_margin” it would basically do its job. If you are not an experienced CRM Developer, who remembers CRM 3.0 or 4.0, just an ordinary .NET Developer, you would probably say, that this code looks ugly. It’s almost like it was taken from some JavaScript library – where are all the types, what is the type of “new_totalsum”, which year is that, are we still in 2017? And yes, you’ll be right – this code simply looks awful, does not allow to use any of great refactoring tools, because it uses some “magic” dictionaries and keys and without access to CRM you would not even have a clue, what are the proper types of the properties. It’s exactly the same story as with ORM – who of us still uses plain ADO .NET and not, for example, Entity Framework? Such approach is much less error prone, testable and also much more friendly for CRM newbies. Code using early bounds may look like that:

Now it’s much better, but there is a significant problem with this code. We assumed that our plugin runs on Account entity. What if we will register it on Contact entity? Well, it will simply crash when we will try to convert our target!

So the basic approach would be to create a switch statement and simply check the entity type:

Ok so now we can register the plugin on multiple entities, but our code looks even worse than the first version, because we have a lot of code that is copy-pasted. Imagine that the operation would be much more complex and would change frequently in time – we would have to change all the switch statements every time.

To overcome this problem we will take advantage of the fact that all early-bound classes are partial, so they can implement any interface we want. Let’s create such interface and some helper partial classes:

It would be great if we could use this like that:

But unfortunately, our “Target” is an entity and we have to somehow convert it to an early bound. I developed a simple utility class that is capable of that:

Simple explanation of the code – during first execution of the plugin, all the early-bound types are cached in a dictionary, that can be later fast accessed from the code (in this case it’s InterfaceMocker). It simply instantiates the early-bound and copies all attributes inside. There is also an extension method that simplifies the usage for Entity. Now our plugin can simply look like that:

And that’s it! This code looks much better then late bound version, provides type-checking for entity attributes and also is very nice when we implement unit tests, as we have an interface to test, not a full early-bound. If our logic is more complex we can create some additional classes that will be dependent upon IEntityWithTotalSum interface, not Account or Opportunity entity.