Dealing with (members of) heterogeneous arrays - powershell

Question

Let's create books
$a = New-Object –TypeName PSObject
$a | Add-Member –MemberType NoteProperty –Name Title –Value "Journey to the West"
$a | Add-Member –MemberType NoteProperty –Name Price –Value 12
$b = New-Object –TypeName PSObject
$b | Add-Member –MemberType NoteProperty –Name Title –Value "Faust"
$b | Add-Member –MemberType NoteProperty –Name Author –Value "Goethe"
$array1 = $a,$b
$array2 = $b,$a
Now let's display these two arrays
PS D:\Developpement\Powershell> $array1
Title Price
----- -----
Journey to the West 12
Faust
PS D:\Developpement\Powershell> $array2
Title Author
----- ------
Faust Goethe
Journey to the West
So as far as I understand this basically means that what powershell consider to be properties of an array are the properties of its first element (in fact that's not even true because if the first element is $null the next one will be considered). Now that also implies that :
if you call Get-Member on the array, you will only get members of the first element
if you call Convert-ToCvs on the array, you will only export property values for properties defined by the first element
etc
I hardly understand the rationals behind that and this behaviour has made it infuriatingly painful for me to work with heterogeneous arrays in powershell.
I'd like to import data from various external sources, process them and then export them to a cvs file. Items are similar but most of them miss some properties unpredictably. Is there any obvious way to handle that in Powershell without reprogramming the wheel?

Answer 1

This is the way it has to be because PowerShell uses pipelines. When you run ex. $array1 | Export-CSV ...., PowerShell starts to write to the CSV-file as soon as the first object arrives. At that point it needs to know what the header will look like as that is the first line in a csv-file. So PowerShell has to assume that the class/properties of the first object represents all the remaining objects in the pipeline. The same goes for Format-Table and similar commands that need to set a style/view before outputting any objects.
The usual workaround to this is to specify the header manually using Select-Object. It will add all missing properties to all objects with a value of $null. This way, all the objects sent to ex. Export-CSV will have all the same properties defined.
To get the header, you need to receive all unique property-names from all objects in your array. Ex.
$array1 |
ForEach-Object { $_.PSObject.Properties} |
Select-Object -ExpandProperty Name -Unique
Title
Price
Author
Then you can specify that as the header using Select-Object -Properties Title,Price,Author before sending the objects to Export-CSV Ex:
$a = New-Object –TypeName PSObject
$a | Add-Member –MemberType NoteProperty –Name Title –Value "Journey to the West"
$a | Add-Member –MemberType NoteProperty –Name Price –Value 12
$b = New-Object –TypeName PSObject
$b | Add-Member –MemberType NoteProperty –Name Title –Value "Faust"
$b | Add-Member –MemberType NoteProperty –Name Author –Value "Goethe"
$array = $a,$b
$AllProperties = $array |
ForEach-Object { $_.PSObject.Properties} |
Select-Object -ExpandProperty Name -Unique
$array | Select-Object -Property $AllProperties | Export-CSV -Path "mycsv.out" -NoTypeInformation
This will create this CSV-file:
"Title","Price","Author"
"Journey to the West","12",
"Faust",,"Goethe"
If you have mulltiple arrays you can combine them like this $array = $array1 + $array2