# Bring existing data in This guide is for measurements you already have, in a spreadsheet, CSV files or a notebook, to turn into sample files: one Markdown file per sample, in a project. Your original data is only read; SampleKit writes new files beside it. The example is the [home page's](../index.md) tensile specimens: a material, a width and a thickness per specimen, and the force read at each strain, which the testing machine wrote to a CSV file per specimen. ## 1. Make the project ```console $ samplekit tensile-tests init --no-model --no-venv --write ``` This writes `.samplekitrc` and an empty `samples/` folder. `--no-model` leaves the model for later, and `--no-venv` makes no Python environment, which reading and writing samples do not need; `init` without them asks ([start a project](start-a-project.md)). ## 2. Declare your properties A measured quantity with a unit is a **property**; a fact without one, the material or the grade, is an **attribute** ([samples and their files](../concepts/samples-and-files.md)). Declare each property in `.samplekitrc`, with its unit and its digits, before writing any sample: ```toml [property.width] unit = "mm" precision = ".2f" [property.thickness] unit = "mm" precision = ".2f" ``` A name `.samplekitrc` declares is then written as a property, with its unit, and so is one the project's model declares, once there is one. A name nothing declares is written as an attribute, and `new` and `set` warn you when that name is given a number. ## 3. Write the samples **One sample** at a time, from the command line: ```console $ samplekit samples new steel-01 material=steel grade=S355 width=12.51 thickness=2.00 --write ``` `width.u=0.01` gives an uncertainty, and `width.readings=12.494,12.528,12.518` repeated measurements. Write decimals with a point, never a comma ([edit samples](edit-samples.md)). **From a CSV file**, a shell loop gives each row to `new`. With `specimens.csv` holding `name,material,grade,width,thickness`, a row per specimen: ```sh tail -n +2 specimens.csv | while IFS=, read -r name material grade width thickness; do samplekit samples new "$name" material="$material" grade="$grade" \ width="$width" thickness="$thickness" --write done ``` From a spreadsheet, save the sheet as CSV first. A field that holds a comma needs a real CSV reader: Python's `csv` module, as below. **A table per sample**, such as each test's curve, is simpler from Python. With `machine/steel-01/curve.csv` holding `strain [%],force [kN]` for each specimen, and its sample written above: ```python import csv from pathlib import Path import samplekit as sk for folder in sorted(Path("machine").iterdir()): sample = sk.Sample(Path("samples") / f"{folder.name}.md") with (folder / "curve.csv").open(newline="") as exported: rows = [{"strain": float(row["strain [%]"]), "force": float(row["force [kN]"])} for row in csv.DictReader(exported)] sample.curve = sk.Table( {"strain": sk.Property(unit="%"), "force": sk.Property(unit="kN")}, index="strain", rows=rows, ) sample.save() ``` Run it with the Python where SampleKit's package is installed ([work from Python](work-from-python.md)). Step 1 made none: `samplekit init`, run again in the project, offers to make it, `.venv/`, or make it yourself ([the Python package](install.md#the-python-package)). The CSV files stay where they are, and `[collection] files = ["machine/{name}/*"]` keeps them attached to their sample ([files](../index.md#files)). **By hand**, write one file in a text editor, then the next ones from it: `new --keep` takes the fields it names from a sample, and nothing else: ```console $ samplekit samples new steel-02 --keep samples/steel-01.md material,grade --write ``` **In the terminal interface**, `N` on the list makes a new sample beside the one under the cursor: its name, then `field=value` for each value, and `keep:material,grade` to take fields from the one under the cursor. `e` on its screen fills a value, and `A` adds an attribute; each change is shown before `y` writes it ([work in the terminal interface](terminal-interface.md)). ## 4. Check `list fields` shows each name under *properties* or *attributes*: a property listed as an attribute was not declared in step 2. `validate` reports what is wrong in a file, and changes nothing. Then compare a few files with your original data by eye before trusting the rest: ```console $ samplekit list fields properties (2) width thickness attributes (4) material grade direction tested always name path filename project state tags $ samplekit validate 3 samples · 0 defects · 0 notes ``` ## Next - [Write a model](first-model.md): what your formulas compute from these values. - Data from several series, or several people: [several projects](several-projects.md).