Appendix F — Glossary

Terms and acronyms used in this guide

Keywords

snowflake, R, RStudio, Posit, VS Code, workspace notebooks, snowflakeR, RSnowflake, mlops

Alphabetical reference for readers new to Snowflake or MLOps. See also Snowflake Platform Primer.

F.1 A

ADBC (Arrow Database Connectivity)
Open standard for database access using Apache Arrow columnar memory — comparable role to ODBC/JDBC, optimized for analytics bulk transfer. RSnowflake optionally uses the Snowflake ADBC driver (adbcsnowflake R package) for large reads/writes as Arrow streams; everyday SQL still uses the SQL REST API (JSON). Typical install: options(RSnowflake.backend = "adbc"). See Platform Primer — ADBC, SQL API, RSnowflake: Query.

Account
Your Snowflake tenant identifier (e.g. xy12345.us-east-1). Used in connection strings and connections.toml.

Account URL / Host
HTTPS endpoint for API calls. Workspace may use an internal gateway host — RSnowflake detects this automatically.

F.2 C

Compute pool
SPCS resource pool that runs containers (inference services, R workers). Distinct from a SQL warehouse.

connections.toml
Standard Snowflake connection profiles file (~/.snowflake/connections.toml). Shared by RSnowflake, snowflakeR, Python connector, and CLI.

CRE (Custom Runtime Image)
Pre-built container image for Workspace/ML Jobs with R and packages baked in — faster than cold bootstrap.

CustomModel
Python Model Registry wrapper type. snowflakeR registers R models by wrapping them in a CustomModel that calls R via rpy2 at inference time.

F.3 D

Dataset (ML)
Immutable, versioned training snapshot in Snowflake ML — often built from Feature Store views.

DBI
R standard database interface. RSnowflake implements DBI over the SQL API.

dbplyr
R package translating dplyr verbs to SQL. Works with RSnowflake connections.

Dynamic Table (DT)
Snowflake object that incrementally materializes query results — common backing store for Feature Views.

F.4 E

EAI (External Access Integration)
Snowflake network object allowing outbound HTTPS from notebooks/containers (CRAN, GitHub, conda/pip indexes).

Entity (Feature Store)
Definition of business keys (e.g. CUSTOMER_ID) used to join features for training and serving.

F.5 F

Feature Store
Snowflake ML service for governed feature definitions, point-in-time joins, and datasets.

Feature View
Versioned feature definition (often backed by a Dynamic Table or View) registered in Feature Store.

F.6 G

GET / PUT (SQL)
Client-side stage file commands. Not supported by the SQL REST API — use stage volume mounts in SPCS instead.

F.7 I

IPython / Jupyter magic
Cell prefix like %%R that routes cell body to another language via rpy2 in Workspace.

IRkernel
CRAN package that registers a native R Jupyter kernel (common on desktop Jupyter). Workspace Notebooks today use a Python kernel + %%R magic instead — see Appendix G.

F.8 L

Lineage (ML)
Traceability from source tables → features → datasets → models → predictions.

F.9 M

ML Container Runtime
Managed environment for Workspace notebooks and ML Jobs (Python kernel; R added via bootstrap).

ML Job
Batch execution on container runtime — non-interactive alternative to notebooks.

micromamba
Fast conda-compatible installer used by setup_notebook() to provision R in Workspace.

Model Registry
Snowflake ML service for versioning, deploying, and serving models.

ModelVersion
Handle to one registered version of a model (sfr_log_model() return value).

F.10 O

OAuth / Workspace token
In Workspace, the notebook session token authenticates RSnowflake/snowflakeR without storing passwords in code.

F.11 P

PAT (Programmatic Access Token)
Snowflake token for API access — alternative to key-pair for automation.

Point-in-time correct
Joining features as they were known at a historical timestamp — prevents training leakage.

Posit Workbench Native App
Posit/RStudio IDE deployed as a Snowflake Native Application — in-account R development.

F.12 R

REST / SQL API
HTTPS interface for submitting SQL without JDBC/ODBC. RSnowflake’s default transport.

reticulate
R package bridging to Python. snowflakeR uses it to call snowflake-ml-python.

Role
Snowflake RBAC identity controlling access to warehouses, databases, and integrations.

rpy2
Python↔︎R bridge: embeds R in Python process (%%R cells) and serves R models in CustomModel.

RSnowflake
DBI driver for Snowflake using SQL REST API (+ optional ADBC).

F.13 S

Snowflake ML
Umbrella for Feature Store, Model Registry, experiments, monitoring (Python SDK + snowflakeR).

Snowpark
Snowflake’s programmatic data framework (Python, Java, Scala). Underlies many ML operations.

Snowsight
Snowflake web UI (SQL worksheets, admin, Workspace).

Spine
Table of entity keys + event timestamps (+ labels) used to request point-in-time feature values.

Stage
Internal file storage location in Snowflake (@db.schema.stage/path).

SPCS (Snowpark Container Services)
Run Docker containers inside Snowflake — inference endpoints, custom R servers, parallel workers.

snowflakeR
R API for Snowflake ML platform (Feature Store, Registry, doSnowflake, …).

SQL API
Snowflake REST endpoints (/api/v2/statements) for executing SQL over HTTPS. See chapter 11.

F.14 T

Task
Scheduled or triggered Snowflake job — often SQL — used in pipelines and doSnowflake backends.

Tarball (R package)
Pre-built .tar.gz R package archive. Required for fast Workspace installs of snowflakeR/RSnowflake.

F.15 W

Warehouse (virtual warehouse)
Elastic SQL compute cluster. Used for queries, materializations, and warehouse-side inference — not arbitrary R code.

WebR
R compiled to WebAssembly for browser execution. Can pair with RSnowflake + SQL API where drivers are unavailable.

Workspace
Snowsight development area with notebooks, files, and Git integration.

Workspace Notebook
Interactive notebook running on ML Container Runtime inside Snowflake.

/filesystem/
FUSE mount where Workspace stages git project files; path includes a content hash. See Appendix G.

F.16 Package names (this guide)

Name	What it is
snowflake-notebook-multilang	Bootstrap toolkit (`setup_notebook()`, `sfnb_setup.py`)
RSnowflake	DBI driver via SQL API
snowflakeR	R API for Snowflake ML platform

--- title: "Glossary" subtitle: "Terms and acronyms used in this guide" --- Alphabetical reference for readers new to Snowflake or MLOps. See also [Snowflake Platform Primer](../../01_snowflake_platform/index.qmd). ## A **ADBC (Arrow Database Connectivity)** Open standard for database access using **Apache Arrow** columnar memory — comparable role to ODBC/JDBC, optimized for analytics bulk transfer. RSnowflake optionally uses the Snowflake ADBC driver (`adbcsnowflake` R package) for large **reads/writes** as Arrow streams; everyday SQL still uses the **SQL REST API** (JSON). Typical install: `options(RSnowflake.backend = "adbc")`. See [Platform Primer — ADBC](../../01_snowflake_platform/index.qmd#sec-map), [SQL API](../../11_rsnowflake_sql_api/index.qmd#sec-arrow), [RSnowflake: Query](../../13_rsnowflake_query/index.qmd#sec-arrow). **Account** Your Snowflake tenant identifier (e.g. `xy12345.us-east-1`). Used in connection strings and `connections.toml`. **Account URL / Host** HTTPS endpoint for API calls. Workspace may use an internal gateway host — RSnowflake detects this automatically. ## C **Compute pool** SPCS resource pool that runs containers (inference services, R workers). Distinct from a SQL **warehouse**. **connections.toml** Standard Snowflake connection profiles file (`~/.snowflake/connections.toml`). Shared by RSnowflake, snowflakeR, Python connector, and CLI. **CRE (Custom Runtime Image)** Pre-built container image for Workspace/ML Jobs with R and packages baked in — faster than cold bootstrap. **CustomModel** Python Model Registry wrapper type. snowflakeR registers R models by wrapping them in a `CustomModel` that calls R via rpy2 at inference time. ## D **Dataset (ML)** Immutable, versioned training snapshot in Snowflake ML — often built from Feature Store views. **DBI** R standard database interface. RSnowflake implements DBI over the SQL API. **dbplyr** R package translating dplyr verbs to SQL. Works with RSnowflake connections. **Dynamic Table (DT)** Snowflake object that incrementally materializes query results — common backing store for Feature Views. ## E **EAI (External Access Integration)** Snowflake network object allowing outbound HTTPS from notebooks/containers (CRAN, GitHub, conda/pip indexes). **Entity (Feature Store)** Definition of business keys (e.g. `CUSTOMER_ID`) used to join features for training and serving. ## F **Feature Store** Snowflake ML service for governed feature definitions, point-in-time joins, and datasets. **Feature View** Versioned feature definition (often backed by a Dynamic Table or View) registered in Feature Store. ## G **GET / PUT (SQL)** Client-side stage file commands. **Not supported** by the SQL REST API — use stage volume mounts in SPCS instead. ## I **IPython / Jupyter magic** Cell prefix like `%%R` that routes cell body to another language via rpy2 in Workspace. **IRkernel** CRAN package that registers a **native R Jupyter kernel** (common on desktop Jupyter). Workspace Notebooks today use a **Python kernel + `%%R` magic** instead — see [Appendix G](../G_workspace_container_internals/index.qmd#sec-ws-irkernel-vs-magic). ## L **Lineage (ML)** Traceability from source tables → features → datasets → models → predictions. ## M **ML Container Runtime** Managed environment for Workspace notebooks and ML Jobs (Python kernel; R added via bootstrap). **ML Job** Batch execution on container runtime — non-interactive alternative to notebooks. **micromamba** Fast conda-compatible installer used by `setup_notebook()` to provision R in Workspace. **Model Registry** Snowflake ML service for versioning, deploying, and serving models. **ModelVersion** Handle to one registered version of a model (`sfr_log_model()` return value). ## O **OAuth / Workspace token** In Workspace, the notebook session token authenticates RSnowflake/snowflakeR without storing passwords in code. ## P **PAT (Programmatic Access Token)** Snowflake token for API access — alternative to key-pair for automation. **Point-in-time correct** Joining features as they were known at a historical timestamp — prevents training leakage. **Posit Workbench Native App** Posit/RStudio IDE deployed as a Snowflake Native Application — in-account R development. ## R **REST / SQL API** HTTPS interface for submitting SQL without JDBC/ODBC. RSnowflake's default transport. **reticulate** R package bridging to Python. snowflakeR uses it to call `snowflake-ml-python`. **Role** Snowflake RBAC identity controlling access to warehouses, databases, and integrations. **rpy2** Python↔R bridge: embeds R in Python process (`%%R` cells) and serves R models in CustomModel. **RSnowflake** DBI driver for Snowflake using SQL REST API (+ optional ADBC). ## S **Snowflake ML** Umbrella for Feature Store, Model Registry, experiments, monitoring (Python SDK + snowflakeR). **Snowpark** Snowflake's programmatic data framework (Python, Java, Scala). Underlies many ML operations. **Snowsight** Snowflake web UI (SQL worksheets, admin, Workspace). **Spine** Table of entity keys + event timestamps (+ labels) used to request point-in-time feature values. **Stage** Internal file storage location in Snowflake (`@db.schema.stage/path`). **SPCS (Snowpark Container Services)** Run Docker containers inside Snowflake — inference endpoints, custom R servers, parallel workers. **snowflakeR** R API for Snowflake ML platform (Feature Store, Registry, doSnowflake, …). **SQL API** Snowflake REST endpoints (`/api/v2/statements`) for executing SQL over HTTPS. See [chapter 11](../../11_rsnowflake_sql_api/index.qmd). ## T **Task** Scheduled or triggered Snowflake job — often SQL — used in pipelines and doSnowflake backends. **Tarball (R package)** Pre-built `.tar.gz` R package archive. Required for fast Workspace installs of snowflakeR/RSnowflake. ## W **Warehouse (virtual warehouse)** Elastic SQL compute cluster. Used for queries, materializations, and warehouse-side inference — not arbitrary R code. **WebR** R compiled to WebAssembly for browser execution. Can pair with RSnowflake + SQL API where drivers are unavailable. **Workspace** Snowsight development area with notebooks, files, and Git integration. **Workspace Notebook** Interactive notebook running on ML Container Runtime inside Snowflake. **/filesystem/** FUSE mount where Workspace stages git project files; path includes a content hash. See [Appendix G](../G_workspace_container_internals/index.qmd). ## Package names (this guide) | Name | What it is | |------|------------| | **snowflake-notebook-multilang** | Bootstrap toolkit (`setup_notebook()`, `sfnb_setup.py`) | | **RSnowflake** | DBI driver via SQL API | | **snowflakeR** | R API for Snowflake ML platform |