Skip to main content

Serialize data with PowerShell

imageCurrently I am working on a big new module. In this module, I need to persist data to disk and reprocess them at some point even if the module/PowerShell session was closed. I needed to serialize objects and save them to disk. It needed to be very efficient to be able to support a high volume of objects. Hence I decided to turn this serializer into a module called HashData.



Other Serializing methods

In PowerShell we have several possibilities to serialize objects. There are two cmdlets you can use which are built in:
  • Export-CliXml
  • ConvertTo-JSON

Both are excellent options if you do not care about the size of the file. In my case I needed something lean and mean in terms of the size on disk for the serialized object. Lets do some tests to compare the different types:


(Hashdata.Object.ps1)

You might be curious why I do not use the Export-CliXML cmdlet and just use the [System.Management.Automation.PSSerializer]::Serialize static method. The static method will generate the same xml, however we do not need to read back the content of the file the cmdlet creates. 

If we compare the length of the string we get this:

image

As you can see, the XML serialization is very bloated with metadata, however the JSON serialization is much better. The winner is the HashData module with a 30% smaller size compared to a JSON string.


HashData module

Currently the module implements these cmdlets:

  • Assert-ScriptString
  • ConvertTo-HashString
  • ConvertTo-Hashtable
  • Export-HashData    
  • Import-HashData    
  • New-Date 

Like for the Import-XMLCli and Export-XMLCli, the logic for serialization and deserialization is implemented in Import-HashData and Export-HashData. I chose to also include and export from the module the helper functions ConvertTo-Hashtable and ConvertTo-HashString. Those could be useful in other scenarios as well. The New-Date function is probably my smallest function I have ever published. It purpose is to be able to convert datetime objects on deserializing objects.

Lets inspect the object we created above and look at it’s string representation:

(HashTextObject.ps1)

As you can see, the datetime object are converted to a [long] ticks value, which the function New-Date converts to a datetime object on deserialize.


Currently implemented property-types

In this version, your object may have properties of the following type:

  • String
  • Integer
  • Boolean
  • Double
  • DateTime
  • Array of String
  • Array of Integers


Currently supported and tested object depth is 1. That might change in the future. You may pipe or supply an array of PSCustomObject to the Export-HashData function.

I have deliberately chosen not to convert the objects from Import-Hashdata to PSCustomObject in this release. Depending on feedback and the need, I will consider adding this at a later stage.


Security

The Assert-ScriptString function is a security boundary and is implemented and used in the Import-Hashdata function. The reason for that is, when you serialize a object as an hashtable string, you are in essence generating a script file which in this instance will behave like a scriptblock. When you Import something, that is invoking a scriptblock, the Assert-ScriptString will make sure nothing evil will ever execute. The only function allowed in the serialized object currently, is the New-Date function.

The Import-HashData function has a switch parameter (UnsafeMode) that lets you override this security feature. Use it with care.


PowershellGallery and GitHub

The module is published to the PowershellGallery https://www.powershellgallery.com/packages/hashdata and here is the link to the GitHub repro https://github.com/torgro/HashData.

Please reach out to me on twitter or leave a comment. I love feedback both good and bad.


Cheers


Tore

Comments

Popular posts from this blog

Toying with audio in powershell

Controlling mute/unmute and the volume on you computer with powershell. Add-Type -TypeDefinition @' using System.Runtime.InteropServices; [Guid("5CDF2C82-841E-4546-9722-0CF74078229A"), InterfaceType(ComInterfaceType.InterfaceIsIUnknown)] interface IAudioEndpointVolume { // f(), g(), ... are unused COM method slots. Define these if you care int f(); int g(); int h(); int i(); int SetMasterVolumeLevelScalar(float fLevel, System.Guid pguidEventContext); int j(); int GetMasterVolumeLevelScalar(out float pfLevel); int k(); int l(); int m(); int n(); int SetMute([MarshalAs(UnmanagedType.Bool)] bool bMute, System.Guid pguidEventContext); int GetMute(out bool pbMute); } [Guid("D666063F-1587-4E43-81F1-B948E807363F"), InterfaceType(ComInterfaceType.InterfaceIsIUnknown)] interface IMMDevice { int Activate(ref System.Guid id, int clsCtx, int activationParams, out IAudioEndpointVolume aev); } [Guid("A95664D2-9614-4F35-A746-DE8DB63617E6"), Inte

Creating Menus in Powershell

I have created another Powershell module. This time it is about Console Menus you can use to ease the usage for members of your oranization. It is available on GitHub and published to the PowershellGallery . It is called cliMenu. Puppies This is a Controller module. It uses Write-Host to create a Menu in the console. Some of you may recall that using Write-Host is bad practice. Controller scripts and modules are the exception to this rule. In addition with WMF5 Write-Host writes to the Information stream in Powershell, so it really does not matter anymore. Design goal I have seen to many crappy menus that is a mixture of controller script and business logic. It is in essence a wild west out there, hence my ultimate goal is to create something that makes it as easy as possible to create a menu and change the way it looks. Make it easy to build Menus and change them Make it as "declarative" as possible Menus The module supports multiple Men