# How to parse node data JSONs from the ncov workflow (or any Nextstrain workflow)

**URL:** <https://nextstrain.discourse.group/t/how-to-parse-node-data-jsons-from-the-ncov-workflow-or-any-nextstrain-workflow/847>\
**Category:** General\
**Created:** [November 15, 2021, 7:09pm UTC](https://nextstrain.discourse.group/t/how-to-parse-node-data-jsons-from-the-ncov-workflow-or-any-nextstrain-workflow/847 "2021-11-15T19:09:32Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![jlhudd](https://yyz1.discourse-cdn.com/flex031/user_avatar/nextstrain.discourse.group/jlhudd/32/583_2.png) [@jlhudd](https://nextstrain.discourse.group/u/jlhudd)\
**Post date:** [November 15, 2021, 7:09pm UTC](https://nextstrain.discourse.group/t/how-to-parse-node-data-jsons-from-the-ncov-workflow-or-any-nextstrain-workflow/847/1 "2021-11-15T19:09:32Z")

</div>

We recently got a question from a Nextstrain user about the best way to parse the JSON output of the ncov workflow. For example, the ncov workflow runs `augur traits` by default to infer the ancestral country for each internal node of the tree. The workflow writes these data to files we refer to as “[node data JSONs](https://docs.nextstrain.org/projects/augur/en/stable/usage/json_format.html)”. We don’t have a how-to guide yet for how to parse these files, but the following code examples show how to get started parsing these files.

Each node data JSON from Augur contains a dictionary with a `nodes` key that indexes another dictionary by strain (or internal node) name. From a Python terminal window the following steps will load ancestral trait values into memory (showing an example based on our Nextstrain build for Europe):

```python
>>> from augur.utils import read_node_data
>>> traits = read_node_data("results/europe/traits.json")
>>> traits_by_node_name = traits["nodes"]

```

The traits JSON also stores information about the discrete trait analysis model in a top level key named `model`. This key points to a dictionary of all discrete trait columns that were inferred by augur traits. In the example data above, the only trait inferred was “country”. Indexing the traits data by `models` and `country` gives us a dictionary with these keys:

```python
>>> traits["models"]["country"].keys()
dict_keys(['alphabet', 'equilibrium_probabilities', 'rate', 'transition_matrix'])

```

Note that Augur’s [`read_node_data` function](https://docs.nextstrain.org/projects/augur/en/stable/api/augur.utils.html#augur.utils.read_node_data) accepts multiple JSON filenames as input; it will combine the data from these multiple files into a single dictionary.

You can use a similar approach to extract these values from R with the jsonlite library like so:

```auto
traits <- jsonlite::fromJSON("results/europe/traits.json")

```

A while back, I wrote [a script to parse these JSON files into TSV/CSV files](https://github.com/blab/flu-forecasting/blob/master/scripts/node_data_to_table.py). This script is a little out-of-date, so [I’ve posted a simpler, updated version on GitHub as a gist](https://gist.github.com/huddlej/1fd00f8410c40826a47ec07e2b99c7bb) with instructions on how to use it. Once you have [Augur installed](https://docs.nextstrain.org/projects/augur/en/stable/installation/installation.html), you can run the script like so:

```bash
python3 node_data_to_table.py \
  --tree results/europe/tree.nwk \
  --jsons results/europe/traits.json \
  --include-internal-nodes \
  --annotations build=europe \
  --output traits.tsv

```

Run the script with the `-h` flag to get more details about what the different arguments do.

```bash
python3 node_data_to_table.py -h

```