Files and JSON

Use file ports for persistent artifacts and specialized formats. Use JSON ports for small structured values that should be consumed as an object rather than a table.

Read-only file inputs

input_file returns a path inside the pipeline runtime. Treat that path as read-only.

Python

from pathlib import Path

import flasktrack as ft

source = Path(ft.input_file("sequence"))
sequence_text = source.read_text(encoding="utf-8")

R

source("/opt/flasktrack/flasktrack.R")

sequence_path <- ft_input_file("sequence")
sequence_text <- paste(readLines(sequence_path, warn = FALSE), collapse = "\n")

Create a separate result and publish it through the declared output port. Do not overwrite or rename the input.

Publish an existing file

Python

ft.output_file("report", report_path, "application/pdf")

R

ft_output_file("report", report_path, "application/pdf")

Supplying an accurate content type helps FlaskTrack and downstream destinations handle the file correctly.

CSV and Parquet

CSV and Parquet helpers read or publish file ports. They do not replace Arrow table ports between processing blocks.

Python

imported = ft.input_csv("uploaded_csv")
ft.output_parquet("archive", imported, compression="snappy")

R

imported <- ft_input_csv("uploaded_csv")
ft_output_parquet("archive", imported, compression = "snappy")

JSON configuration

Validate required keys and expected value types before using configuration.

import flasktrack as ft

config = ft.input_json("config")
if not isinstance(config, dict):
    raise ValueError("config must be a JSON object")
if "minimum_value" not in config:
    raise ValueError("config is missing minimum_value")

minimum_value = float(config["minimum_value"])

When publishing JSON, convert library-specific scalar objects to ordinary Python or R values first. Python JSON output rejects NaN and infinity; decide whether those values should become null, a string, or an error before publishing.

Text and binary files

Use output_text for generated plain text and output_bytes for in-memory binary content. Both require a declared file output.

ft.output_text("summary_text", "Processing completed\n")
ft.output_bytes(
    "preview",
    png_bytes,
    extension="png",
    content_type="image/png",
)