Add Find Duplicate Records In Table Without Unique Id as a postgres til

README.md

@@ -10,7 +10,7 @@ pairing with smart people at Hashrocket.
 
 For a steady stream of TILs, [sign up for my newsletter](https://tinyletter.com/jbranchaud).
 
-_1124 TILs and counting..._
+_1125 TILs and counting..._
 
 ---
 
@@ -558,6 +558,7 @@ _1124 TILs and counting..._
 - [Escaping String Literals With Dollar Quoting](postgres/escaping-string-literals-with-dollar-quoting.md)
 - [Export Query Results To A CSV](postgres/export-query-results-to-a-csv.md)
 - [Extracting Nested JSON Data](postgres/extracting-nested-json-data.md)
+- [Find Duplicate Records In Table Without Unique Id](postgres/find-duplicate-records-in-table-without-unique-id.md)
 - [Find Records That Contain Duplicate Values](postgres/find-records-that-contain-duplicate-values.md)
 - [Find Records That Have Multiple Associated Records](postgres/find-records-that-have-multiple-associated-records.md)
 - [Find The Data Directory](postgres/find-the-data-directory.md)

postgres/find-duplicate-records-in-table-without-unique-id.md

@@ -0,0 +1,45 @@
+# Find Duplicate Records In Table Without Unique Id
+
+I recently came across a couple methods for listing out instances of duplicate
+records in a table where the table doesn't have an explicit unique identifier.
+Here is [a post](find-records-that-contain-duplicate-values.md) that explains
+how to do this when a unique identifier is present.
+
+If the table doesn't have an explicit primary key or other uniquely identifying
+value, then we'll have to get some help from [PostgreSQL's internal system
+columns](https://www.postgresql.org/docs/current/ddl-system-columns.html) —
+namely the `ctid`.
+
+The `ctid` is:
+
+> The physical location of the row version within its table.
+
+Let's use the example of the `mailing_list` table with potential duplicate
+`email` values.
+
+Here is the [first approach](https://stackoverflow.com/a/26773018/535590):
+
+```sql
+delete from mailing_list
+where ctid not in (
+  select min(ctid)
+  from mailing_list
+  group by email
+);
+```
+
+This uses a subquery to find the first occurrence of every unique email and
+then deletes the rest. The `ctid` is the unique value that we can call the
+`min` aggregate on.
+
+A [second approach](https://stackoverflow.com/a/46775289/535590):
+
+```sql
+delete from mailing_list ml1
+  using mailing_list ml2
+where ml1.ctid < ml2.ctid
+  and ml1.email = ml2.email;
+```
+
+This uses `delete using` to join the table against itself as a cartesian
+product to compare every entry to every other entry.