While metaframe supports the ability to run ETL jobs and automatically document tables, you can also use it fully manually if you so desire to document whatever you'd like -- tables, ML models, dashboards. The actual base idea behind mf
is pretty flexible.
If you want to start documenting a new dataset, run the following
mf new <TABLE_NAME>
replacing <TABLE_NAME>
with your table's name. This will create a new metaframe table stub and open up a <TABLE_NAME>.docs.md
file, which will allow you to start writing documentation for this table immediately. The metadata will then be searchable using mf
, as usual.
We follow the naming convention database/cluster.schema.table
(and drop the cluster
if there is no cluster or catalog), which, if you want to maintain compatibility with future ETL jobs, we recommend you follow as well.
The mf
command searches over your ~/.metaframe/metadata
directory for two kinds of files:
table_name.md
files These contain the more or less unchanging properties about your dataset. Because these are not by default visible or editable bymf
, you probably don't want to store any frequently changing data here, as this will be overwritten by any future ETL jobs. If you do, however, you can runmf --all
to show and edit all raw files.table_name.docs.md
files These contain any custom docs you want to add regarding the dataset. This is the file that you enter when you pressenter
in themf
window, so this is a great place to store quick notes or anything you need to be easily editable.