TheCOPY INTOcommand supports loading structured and semi-structured data from various Snowflake stages. Supported file formats includeCSV,JSON,Parquet,Avro,ORC, and others. Each format integrates with Snowflake’s file format definitions, enabling configuration for delimiters, compression, headers, and parsing rules.

CSVis widely used for tabular datasets and supports flexible parsing via FILE_FORMAT options such as FIELD_DELIMITER, SKIP_HEADER, and NULL_IF.

JSONis loaded natively into VARIANT columns, allowing nested structures to be queried immediately through path notation.

Parquet, a high-efficiency columnar format, is optimized for analytic workloads and provides excellent compression and schema evolution benefits.

XLS is not supported, as Snowflake does not support Excel spreadsheets directly. XLS/XLSX files must be converted to CSV or Parquet before loading.

The COPY INTO architecture ensures performant ingestion across multi-node compute clusters.

====================================================

Standard permanent tables in Snowflake have a default1-day Time Travel retention period. During this period, users may query historical data, restore dropped tables, or clone past states. Higher editions allow extending this period up to 90 days. Temporary and transient tables have 0-day retention by default.

====================================================

Snowflake processes queries usingMassively Parallel Processing (MPP)compute clusters, deployed as virtual warehouses. Each warehouse consists of multiple compute nodes working in parallel to execute queries efficiently. When a query is submitted, Snowflake distributes tasks across nodes, processes data subsets concurrently, and aggregates results. This architecture enables high performance, scalability, and the ability to handle complex analytical workloads. While Snowflake does incorporate elements of shared-disk storage, query execution itself depends on MPP compute clusters. Options such as third-party connectors or storage optimization do not represent the core query processing mechanism.

=======================================

Schemas serve as logical containers within a database, grouping tables, views, file formats, functions, and other objects.

Roles control access, not organization.

Tables store data but do not group objects.

Stages store files or point to external storage; they are not organizational containers.

==================

In Snowflake Notebooks, a Markdown cell is a non-executable cell type used for formatted text. It allows the user to format text using Markdown syntax, including headings, lists, tables, emphasis, inline code, and links. This makes it possible to document the analysis, describe steps, and provide commentary alongside SQL and Python cells, improving clarity and collaboration.

Markdown cells do not execute code in “more than one computer language”; they are not code cells at all. Notebook cells are not nested; each cell exists as a separate element in the notebook. There is no concept of “older data” associated specifically with Markdown cells; they simply store text content defined by the user.

==================

A red checkmark in Snowflake Notebooks indicates that the cell encountered anerrorduring execution. This visual indicator helps users quickly identify issues so they can inspect the error message and adjust code accordingly. Successful execution shows a green checkmark, while running cells show a spinner. Snowflake notebooks do not include a “paused” state for notebook cells.

=======================================

TheCOMPLETEfunction is the only Cortex LLM function where the user can explicitly specify the model name (e.g., Mistral, Snowflake Arctic, Meta Llama models).

SUMMARIZE, SENTIMENT, and TRANSLATE are task-specific functions that use built-in, preselected models anddo not allowmodel selection.

The SENTIMENT function in Snowflake Cortex analyzes text and returns a numerical sentiment score between -1 and 1. This value indicates whether the text expresses negative, positive, or neutral sentiment. It accepts English-language text and uses Snowflake’s managed AI models to infer polarity based on contextual understanding. A negative score represents negative emotion, a positive score represents positive emotion, and values near zero indicate neutrality. It is ideal for customer reviews, feedback analysis, and text mining. TRANSLATE is used for language translation, COMPLETE for text generation, and PARSE_DOCUMENT for extracting text from documents—not sentiment analysis. SENTIMENT provides fast, in-database AI inference without external model hosting or integration.

=======================================

The FILE_FORMAT parameter is the correct and official Snowflake option used with the COPY INTO < TABLE > command to define the structure and characteristics of the source data files. When loading data, Snowflake needs to understand how to interpret the incoming file—its delimiter, compression, data type (CSV, JSON, PARQUET, etc.), and any additional parsing rules. The FILE_FORMAT parameter can reference either a named file format created earlier (CREATE FILE FORMAT) or an inline definition directly within the COPY INTO statement. Parameters like TYPE, FIELD_DELIMITER, SKIP_HEADER, and ESCAPE can be specified within FILE_FORMAT. None of the other options—DATA_TYPE, FILE_TYPE, LOAD_FORMAT—are valid COPY INTO parameters in Snowflake. DATA_TYPE relates to column definitions, FILE_TYPE is not a supported keyword, and LOAD_FORMAT is not recognized by Snowflake. Using FILE_FORMAT ensures proper data ingestion, prevents parsing errors, and enables Snowflake to correctly interpret semi-structured and structured files during bulk load operations.

=======================================

The only valid Snowflake SQL command among the options forcreating a schemais:

CREATE SCHEMA schema_name;

This command creates a schema within the currently active database. Optional modifiers like IF NOT EXISTS or OR REPLACE can be used to control behavior.

Other options:

USE SCHEMAswitches to an existing schema; it does not create one.

USE SCHis invalid because SCH is not a Snowflake keyword.

CREATE SCHis invalid; SCH is not a valid abbreviation.

Thus, the only valid schema creation command isCREATE SCHEMA.

====================================================