mf-translate

mf-translate converts MetricFlow models and metrics to alternative semantic layers. The idea is to lighten the load of parallel running of two semantic layers by enabling automated synchronisation of definitions. The repo also includes mf-compare-query for checking that equivalent queries against different semantic layers return the same result.

Quickstart

1. Install

pip install git+https://github.com/benw-at-birdie/mf-translate.git

2. Translate a model

Translate all the dimensions and metrics from the deliveries semantic model.

cd your_dbt_project_directory/
mf-translate --model deliveries --to-looker-view deliveries_base > looker/deliveries_base.view.lkml

3. Compare semantic layer queries

Setup Looker credentials:

export LOOKERSDK_BASE_URL="https://your.looker.instance:19999"
export LOOKERSDK_CLIENT_ID="your_client_id"
export LOOKERSDK_CLIENT_SECRET="your_client_secret"

Set target Looker model:

export MF_TRANSLATE_LOOKER_MODEL="jaffle_shop"

Note that the model name is case sensitive.

Compare a MetricFlow query results to the equivalent Looker query:

mf-compare-query --to-looker-explore deliveries --metrics deliveries_count --group-by delivery_person_id__full_name

See the mf-compare-query readme for more information.

Supported Semantic Layers

Currently only Looker is supported as a translation destination but there is potential to expand support to Cube.dev and Lightdash. Also, mf-translate only supports one-way translation, it is not possible to translate a LookML model back to MetricFlow for example. Below are the metric types which can currently be translated: -

Metric Type	Looker	Cube.dev	Lightdash
simple	✅
ratio	✅
derived
cumulative
conversion