Datanymizer comparison

[gregwebs]

Datanymizer and Greenmask are both extensible open source DB masking tools that work with PostgreSQL.

Language

Datanymizer is in Rust. greenmask is in Go. The language shouldn't make much difference for the end user. Generally Rust is a bit faster than Go, but there are some differences in architecture that would probably end up dominating run time.
But for someone that wants to contribute custom filters or code changes this matters.

Maintenership

Datanymizer has unmerged and thus unreleased PRs that are quite important. Currently greenmask seems to be maintained enthusiastically.

Subsetting

Datanymizer supports subsetting, but greenmask does not yet support it. See #18

Transform condition

Datanymizer supports giving a transform_condition which specifies a subset of rows that should not be transformed. Greenmask does not yet have this feature.

Filters

Both mostly leverage an underlying faker library. A more thorough comparison would be needed here.

Architecture

Datanymizer intercepts the pg_dump output before it is written to disk as a dump file. When writing templates one has access to the previous and current values of columns as the strings they would be written to in the dump file. This plus working within in a templating language inside YAML can make some transformations very difficult.

I believe that greenmask operates differently here although I am not certain. This may effect the total time to produce the anonymized dump and affects the ergonomics of custom templating.

Altering multiple columns together

For columns that are dependent on the generated output of other columns datanymizer works fine. However, it doesn't have a facility for generating multiple columns at once. The main example of this would be generating an address. In datanymizer the address will be non-sensical if there is a separate column for city, state, street, etc. In greenmask, this should be able to be handled without issue with the TemplateRecord feature.

Templating and custom filters

In Datanymizer one only has access to template language functions, the previous column value, the new column value, and the values of other columns. However, in greenmask one has access to all of the generator and faker functions from the template.