#### 2020-07-13

readwritesqlite is an R package to enhance reading and writing to SQLite databases.

## Starting

The first task after loading readwritesqlite is to create an object of class SQLiteConnection. Below we create one in memory although in general the user will want to specify a path.

library(readwritesqlite)
conn <- rws_connect(":memory:")

## Writing Data

Individual data frames, environments or named lists of data frames can be written to a connection using rws_write().

In the case of a data frame the default table name is the name of the object (an alternative name can be specified using the x_name argument).

rws_write(rws_data, exists = FALSE, conn = conn)
rws_list_tables(conn)
#> [1] "rws_data"

The fact that rws_write() accepts environments means that the user can easily write all the data frames in the current environment to a SQLiteConnection.

a_table <- rws_data[c("date", "logical")]
another_table <- rws_data[c("factor", "ordered")]
not_a_table <- 1
rws_write(environment(), exists = FALSE, all = FALSE, conn = conn)
rws_list_tables(conn)
#> [1] "a_table"       "another_table" "rws_data"

Objects which are not data frames are silently filtered from environments but cause an error with lists.

### Exists

By default the exists argument to the rws_write() function is TRUE. This means that only existing tables (which were presumably created by the database designer with appropriate checks and foreign keys) can be written to. If the user wishes to automatically create new tables (if they don’t exist) before writing then they should set exists = NA. If the user wishes to only write to previously non-existent tables then they should set exists = FALSE.

### Foreign Keys

The rws_write() function ensures that writing the new data frame(s) to the database does not violate foreign keys. If any data does the database is left unchanged.

### Committing Data

By default, if no check or key violations occur the data frame(s) are written to the database. Otherwise an error message is issued and all changes are rolled back. If the user wishes to simply check whether data could be written to a database without actually making any changes then they can call rws_write() with commit = FALSE.

### Deleting Data (and Meta Data)

Meta data is recorded if the user uses rws_write() to write a data frame to an empty table. In order to change or add meta data the user should read the existing data from the table (using rws_read_sqlite()), modify it accordingly and then re-write it using rws_write() with delete = TRUE. The only exception is for factors and ordered factors. If the existing factor levels are already recorded in the meta data then the user can pass data with additional or rearranged levels for factor and with additional levels for ordered factors without deleting the existing data.

### Replacing Data

If writing data violates a unique or primary key an error message is returned and the table is unaltered. The only exception to this is if replace = TRUE in which case existing rows which cause unique or primary key violations are removed.

### Duplicate Data

When passing data frames to rws_write() in the form of an environment or named list, each table must be represented by just one data frame if unique = TRUE (the default). Duplicates are also not permitted if delete = TRUE (because the first data to be written will be overwritten) or exists = FALSE (because the table will exist when the duplicate is written).

### All Data

When passing data frames to rws_write() in the form of an environment or named list, if all = TRUE (the default) and exists is not FALSE then each existing tables must be represented at least once. This option is useful for checking all the tables in a data frame are populated when transferring data from an old to new database.

### Strict

By default strict = TRUE and extra columns in an input data frame cause an error to be thrown. If strict = FALSE the error is replaced by a warning and the additional columns are automatically removed from the data. When writing environments or list of data with exists = TRUE, the strict argument also determines if extra data frames cause an informative error or are automatically removed with a warning.

### Silent

If the user wishes to suppress package specific messages or warnings then they should set silent = TRUE. As the default value is silent = getOption("rws.silent", FALSE) the user can silence the package session-wide with options(silent = TRUE).

Data can be read using rws_read() which either takes a vector of table names or the connection as the first argument. rws_read() returns a named list of tibbles. If the connection is the first argument then the named list consists of all tables in the data base.

tables <- rws_read(conn)
names(tables)
#> [1] "a_table"       "another_table" "rws_data"
tables\$rws_data
#> # A tibble: 3 x 6
#>   logical date       factor ordered posixct             units
#>   <lgl>   <date>     <fct>  <ord>   <dttm>                [m]
#> 1 TRUE    2000-01-01 x      x       2001-01-02 03:04:05  10.0
#> 2 FALSE   2001-02-03 y      y       2006-07-08 09:10:11  11.5
#> 3 NA      NA         <NA>   <NA>    NA                     NA

The table names can of course be manipulated and list2env() used to assign the data frames to the current environment.

names(tables) <- toupper(names(tables))
list2env(tables, environment())

If the user wishes to read a single data frame they can use rws_read_table()

rws_read_table("rws_data", conn = conn)
#> # A tibble: 3 x 6
#>   logical date       factor ordered posixct             units
#>   <lgl>   <date>     <fct>  <ord>   <dttm>                [m]
#> 1 TRUE    2000-01-01 x      x       2001-01-02 03:04:05  10.0
#> 2 FALSE   2001-02-03 y      y       2006-07-08 09:10:11  11.5
#> 3 NA      NA         <NA>   <NA>    NA                     NA

The rws_read_meta() and rws_read_log() allow the user to read the meta and log tables.

rws_read_meta(conn)
#> # A tibble: 10 x 4
#>    TableMeta     ColumnMeta MetaMeta          DescriptionMeta
#>    <chr>         <chr>      <chr>             <chr>
#>  1 ANOTHER_TABLE FACTOR     factor: 'x', 'y'  <NA>
#>  2 ANOTHER_TABLE ORDERED    ordered: 'y', 'x' <NA>
#>  3 A_TABLE       DATE       class: Date       <NA>
#>  4 A_TABLE       LOGICAL    class: logical    <NA>
#>  5 RWS_DATA      DATE       class: Date       <NA>
#>  6 RWS_DATA      FACTOR     factor: 'x', 'y'  <NA>
#>  7 RWS_DATA      LOGICAL    class: logical    <NA>
#>  8 RWS_DATA      ORDERED    ordered: 'y', 'x' <NA>
#>  9 RWS_DATA      POSIXCT    tz: Etc/GMT+8     <NA>
#> 10 RWS_DATA      UNITS      units: m          <NA>
rws_read_log(conn)
#> # A tibble: 6 x 5
#>   DateTimeUTCLog      UserLog TableLog      CommandLog NRowLog
#>   <dttm>              <chr>   <chr>         <chr>        <int>
#> 1 2019-07-07 16:09:59 joe     RWS_DATA      CREATE           0
#> 2 2019-07-07 16:10:00 joe     RWS_DATA      INSERT           3
#> 3 2019-07-07 16:10:00 joe     A_TABLE       CREATE           0
#> 4 2019-07-07 16:10:00 joe     A_TABLE       INSERT           3
#> 5 2019-07-07 16:10:00 joe     ANOTHER_TABLE CREATE           0
#> 6 2019-07-07 16:10:00 joe     ANOTHER_TABLE INSERT           3

## Cleaning Up

It’s good practice to close a connection once you have finished with it.

rws_disconnect(conn)