SQLDataFrame: SQL-backed DataFrame

In Bioconductor/SQLDataFrame: Representation of SQL tables in DataFrame metaphor

SQL-backed DataFrame

Description

Create a SQL-backed DataFrame, where the data are kept on disk until requested. Direct extension classes are SQLiteDataFrame and DuckDBDataFrame.

Usage

SQLDataFrame(path, dbtype = NULL, table = NULL, columns = NULL, nrows = NULL)

Arguments

path	String containing a path to a SQL file.
dbtype	String containing the SQL database type (case insensitive). Supported types are "SQLite" and "DuckDB".
table	String containing the name of SQL table.
columns	Character vector containing the names of columns in a SQL table. If NULL, this is determined from path.
nrows	Integer scalar specifying the number of rows in a SQL table. If NULL, this is determined from path.

Details

The SQLDataFrame is essentially just a DataFrame of SQLColumnVector objects. It is primarily useful for indicating that the in-memory representation is consistent with the underlying SQL file (e.g., no delayed filter/mutate operations have been applied, no data has been added from other files). Thus, users can specialize code paths for a SQLDataFrame to operate directly on the underlying SQL table.

In that vein, operations on a SQLDataFrame may return another SQLDataFrame if the operation does not introduce inconsistencies with the file-backed data. For example, slicing or combining by column will return a SQLDataFrame as the contents of the retained columns are unchanged. In other cases, the SQLDataFrame will collapse to a regular DFrame of SQLColumnVector objects before applying the operation; these are still file-backed but lack the guarantee of file consistency.

Value

A SQLDataFrame where each column is a SQLColumnVector.

Author(s)

Qian Liu

Examples

## Mocking up a file:

### SQLite
tf <- tempfile()
on.exit(unlink(tf))
con <- DBI::dbConnect(RSQLite::SQLite(), tf)
DBI::dbWriteTable(con, "mtcars", mtcars)
DBI::dbDisconnect(con)

### DuckDB
tf1 <- tempfile()
on.exit(unlist(tf1))
con <- DBI::dbConnect(duckdb::duckdb(), tf1)
DBI::dbWriteTable(con, "mtcars", mtcars)
DBI::dbDisconnect(con)

## Creating a SQLite-backed data frame:

df <- SQLDataFrame(tf, dbtype = "SQLite", table = "mtcars")
df1 <- SQLiteDataFrame(tf, "mtcars")
identical(df, df1)

## DuckDB-backed data frame:
df2 <- SQLDataFrame(tf1, dbtype = "duckdb", table = "mtcars")
df3 <- DuckDBDataFrame(tf1, "mtcars")
identical(df2, df3)
## Extraction yields a SQLiteColumnVector:
df$carb

## Some operations preserve the SQLDataFrame:
df[,1:5]
combined <- cbind(df, df)
class(combined)

## ... but most operations collapse to a regular DFrame:
df[1:5,]
combined2 <- cbind(df, some_new_name=df[,1])
class(combined2)

df1 <- df
rownames(df1) <- paste0("row", seq_len(nrow(df1)))
class(df1)

df2 <- df
colnames(df2) <- letters[1:ncol(df2)]
class(df2)

df3 <- df
df3$carb <- mtcars$carb
class(df3)

## Utility functions
path(df)
dbtype(df)
sqltable(df)
dim(df)
names(df)

as.data.frame(df)

Bioconductor/SQLDataFrame documentation built on March 29, 2024, 6:41 p.m.