I have had a few questions around de-duping files within a SharePoint environment recently so I set off to do some research to identify a good solution. Based on past experiences I knew that SharePoint identifies duplicates while performing an index of the content so I expected this would be part of the solution.
Upon starting my journey, I found a couple of threads on various forums where the question has been asked in the past. The first one was “Good De-Dup tools for SharePoint” which had a link to a blog post by Gary Lapointe that offered a PowerShell script that can list every library item in a farm. At first glance this seemed to be neat, but not helpful here.
Next I found a blog post with another handy PowerShell script. This blog post was title Finding Duplicate Documents in SharePoint using PowerShell. I found this script interesting, albeit dangerous. It will iterate through all of your site collections, sites, and libraries, hash each document and compare for duplicates. It however only identifies duplicate documents within the same location. The overhead of running this script is going to be pretty high, and it gets a little risky when you have larger content stores. I would be worried about running this against an environment that has 100s of sites, or large numbers of documents.
Next I found an old MSDN thread named Find duplicate files which had two interesting answers. The first was to query the database (very bad idea) and the second was a response by Paul Galvin that pointed to the duplicates keyword property, and a suggestion to execute a series of alpha wildcard searches with the duplicates keyword. While I have used the duplicates keyword before I had never thought to use it in this context so I set out to give it a try.
As I mentioned at the beginning SharePoint Search does identify duplicates documents. It does this by generating a hash of the document. Unlike the option above where the PowerShell generates a hash, the search hash seems to separate out the meta-data so even items with unique locations, meta-data, and document names can still be identified as identical documents.
When doing some tests though I quickly discovered that the duplicates property requires the full document URL. This means that you would have to execute a recursive search. First you would have to get a list of items to work with, and then you would then need to iterate through each of those items and execute the duplicates search with a query such as duplicates:”[full document url]”.
Conceptually there are two paths forward at this point. The first is to try and obtain a list of all items from SharePoint Search. Unfortunately you cannot get a full list of everything. The best you can do is the lose title search that Paul had suggested. Something like title:”a*” which would return all items with an a in the title. You would then have to go through and do that for all letters and numbers. One extra challenge is that you will be repeatedly processing the same items unless you are using FAST Query Language and have access to the starts-with operator and can do something like title:starts-with(“a”). In addition, since we are only looking for documents, its an extremely good idea to also add in the isdocument:true to your query to ensure that only documents are returned. Overall this is a very inefficient process.
An alternative would be to revisit and extend Gary’s original script to execute the duplicates search for each item. The advantage here is that you would guarantee that you are only executing the duplicates search once for each item which would reduce the total processing and extra output information to be parsed. The other change to Gary’s script would be to change what is written out to the log file since you would only write out the information for items that are identified as duplicates.