Please sign in to comment.
Improve user ID randomization, avoid non-ISBN items
By creating a separate table, we can generate a much more random string to use as a surrogate for the user, rather than directly translating their user ID into a hash or including the timestamp as salt (where the timestamp, as a static value, could be brute-forced). Also avoid generating item entries where we have no ISBNs because we consequently have no matchpoints, either. Signed-off-by: Dan Scott <firstname.lastname@example.org>
- Loading branch information...
Showing with 33 additions and 11 deletions.
|@@ -0,0 +1,13 @@|
|+Evergreen Level 0 Data Extract|
|+Horribly simple approach, good enough for demonstration purposes for getting|
|+data to build a recommendation engine. Just plain old SQL to run against the|
|+You will need to update:|
|+1. Institution IDs from which you want to extract data (we're assuming|
|+ a hierarchy where a parent institution contains the children from|
|+ which you want data)|
|+2. Output file names to something not so hardcoded.|