Importing Data and Dataset

Import and Export

Import reads hundreds of formats — CSV, TSV, Excel, JSON, XML, images, audio — and Export writes them. The format comes from the file extension or a second argument.

ImportString["name,age\nAda,36\nAlan,41", "CSV"]
ImportString["{\"a\": 1, \"b\": [2, 3]}", "RawJSON"]
ExportString[<|"x" -> 1, "y" -> {1, 2}|>, "RawJSON", "Compact" -> True]

{{name, age}, {Ada, 36}, {Alan, 41}}
<|a -> 1, b -> {2, 3}|>
{"x":1,"y":[1,2]}

• ImportString and ExportString work on text instead of files — handy for testing. Import["file.csv"] and Export["file.json", data] work the same way with files.
• Numbers in CSV become numbers automatically.
• "RawJSON" maps JSON objects to associations. (Check JSON with the JSON Formatter if an import fails.)

Dataset

A Dataset wraps a list of associations (or other nested data) with a table display and a query language.

ds = Dataset[{
  <|"name" -> "Ada", "dept" -> "IT", "salary" -> 61000|>,
  <|"name" -> "Bo", "dept" -> "Sales", "salary" -> 52000|>,
  <|"name" -> "Cy", "dept" -> "IT", "salary" -> 70000|>}];

In a notebook, ds displays as a table with the columns name, dept, and salary. For files, Import["staff.csv", "Dataset", HeaderLines -> 1] builds the same structure from a CSV with a header row.

Querying

A query is written as ds[rows, columns] — each argument is an operation applied at that level:

ds[Total, "salary"]
ds[Mean, "salary"]
ds[Select[#salary > 55000 &], "name"] // Normal
ds[SortBy[-#salary &], "name"] // Normal
ds[GroupBy["dept"], Total, "salary"] // Normal
ds[Counts, "dept"] // Normal

183000
61000
{Ada, Cy}
{Cy, Ada, Bo}
<|IT -> 131000, Sales -> 52000|>
<|IT -> 2, Sales -> 1|>

• ds[Total, "salary"] — take the "salary" column, then total it.
• Select[...] and SortBy[...] are operator forms: functions waiting for their data.
• GroupBy["dept"] splits the rows by department; the next arguments apply within each group.
• Queries that return tables come back as Datasets; Normal converts them to ordinary lists and associations (used here so the results can be shown as text).

Adding columns

ds[All, <|#, "monthly" -> #salary/12|> &][All, "monthly"] // Normal

{15250/3, 13000/3, 17500/3}

<|#, "monthly" -> ...|> copies each row's association and adds a key. Exact division gives fractions; use Round or N for readable numbers:

ds[All, Round[#salary/12] &] // Normal

{5083, 4333, 5833}

Beyond CSV

• SemanticImport detects dates, quantities, and entities in text data.
• Import reads Excel sheets ("XLSX"), and DatabaseReference and SQLSelect connect to SQL databases.
• Tabular (in recent versions) is a fast, column-oriented table type for large data.