Ciaren

File output

Write the upstream frame to a file. This is usually the last node in a pipeline — the terminal step that produces a downloadable, versionable dataset.

The single File Output node replaces the old per-format output nodes: pick the format from a dropdown and give it a name.

  1. 1Input
    File Input
  2. 2Clean
    Fill Nulls
  3. 3Transform
    Calculated Column
  4. 4Output
    File Output
    format + name → dataset

type: fileOutput.

Use cases

  • Save a cleaned dataset back out as CSV, TSV, Excel, Parquet, JSON, JSON Lines, or text.
  • Produce a Parquet file for fast re-reads in a later flow.

Configuration

Config keyTypeRequiredDescription
formatenumYescsv · tsv · excel · parquet · json · jsonl · text
dataset_namestringYesName for the produced dataset/file

The result is written to the run's output location and registered as a dataset you can preview, download, or feed into another flow.

Legacy outputs

Flows built before the consolidation may still contain the old csvOutput / excelOutput / parquetOutput nodes. They keep working, but new flows use the unified File Output node (the old ones are hidden from the palette).

Generated Python code

These separate flows read source.csv and write result with the selected format extension (dataset_name: result). Put your original CSV beside the script and choose one example. The seven examples use the same source, each with its own destination format. The exported script creates a local file, not a registered Ciaren dataset or run. JSON stores records and JSONL stores one record per line; indentation may differ between engines. Text writes headerless, tab-separated rows without column types; use Parquet to read a table back with its types. These are complete eager exports with imports; use an environment with Ciaren and its file-format dependencies installed.

pandas

CSV

python
import pandas as pd

df_source = pd.read_csv('source.csv')
df_source.to_csv('result.csv', index=False)

TSV

python
import pandas as pd

df_source = pd.read_csv('source.csv')
df_source.to_csv('result.tsv', index=False, sep='\t')

EXCEL

python
import pandas as pd

df_source = pd.read_csv('source.csv')
df_source.to_excel('result.xlsx', index=False)

PARQUET

python
import pandas as pd

df_source = pd.read_csv('source.csv')
df_source.to_parquet('result.parquet', index=False)

JSON

python
import pandas as pd

df_source = pd.read_csv('source.csv')
df_source.to_json('result.json', orient='records', indent=2)

JSONL

python
import pandas as pd

df_source = pd.read_csv('source.csv')
df_source.to_json('result.jsonl', orient='records', lines=True)

TEXT

python
import pandas as pd

df_source = pd.read_csv('source.csv')
df_source.to_csv('result.txt', index=False, header=False, sep='\t')

Polars

CSV

python
import polars as pl

df_source = pl.read_csv('source.csv')
df_source.write_csv('result.csv')

TSV

python
import polars as pl

df_source = pl.read_csv('source.csv')
df_source.write_csv('result.tsv', separator='\t')

EXCEL

python
import polars as pl

df_source = pl.read_csv('source.csv')
df_source.write_excel('result.xlsx')

PARQUET

python
import polars as pl

df_source = pl.read_csv('source.csv')
df_source.write_parquet('result.parquet')

JSON

python
import polars as pl

df_source = pl.read_csv('source.csv')
df_source.write_json('result.json')

JSONL

python
import polars as pl

df_source = pl.read_csv('source.csv')
df_source.write_ndjson('result.jsonl')

TEXT

python
import polars as pl

df_source = pl.read_csv('source.csv')
df_source.write_csv('result.txt', include_header=False, separator='\t')

Tips & common mistakes

  • One terminal node per branch. Each output writes the frame reaching it; use multiple output nodes to materialize intermediate branches.
  • Parquet preserves dtypes. Prefer it over CSV when types (datetimes, integers, categoricals) matter downstream.
  • Text output writes headerless, tab-separated records with CSV-style quoting. Values containing tabs or newlines are escaped, so a record may span physical lines. Text Input reads raw lines and does not reverse this encoding; use TSV or Parquet when you need to read the table back.

See also