character: a database table name accessible from
the connected DSN. If missing, the name of dat.
index
character. Name(s) of index column(s) to be used.
append
logical. Should data be appended to an existing table?
rownames
either logical or character. If logical, save the row
names as the first column rownames in the table? If
character, the column name under which to save the rownames.
colnames
logical: save column names as the first row of table?
verbose
display statements as they are sent to the server?
safer
logical. If true, create a non-existing table but only
allow appends to an existing table. If false, allow sqlSave
to attempt to delete all the rows of an existing table, or to drop it.
addPK
logical. Should rownames (if included) be specified as a
primary key?
typeInfo
optional list of DBMS datatypes. Should have elements
named "character", "double" and "integer".
varTypes
an optional named character vector giving the DBMSs
datatypes to be used for some (or all) of the columns if a table is
to be created.
fast
logical. If false, write data a row at a time. If true,
use a parametrized INSERT INTO or UPDATE query to
write all the data in one operation.
test
logical: if TRUE show what would be done, only.
nastring
optional character string to be used for writing
NAs to the database. See ‘Details’.
Details
sqlSave saves the data frame dat in the table
tablename. If the table exists and has the appropriate
structure it is used, or else it is created anew. If a new table is
created, column names are remapped by removing any characters which
are not alphanumeric or _, and the types are selected by
consulting arguments varTypes and typeInfo, then looking
the driver up in the database used by getSqlTypeInfo or
failing that by interrogating sqlTypeInfo.
If rownames = TRUE the first column of the table will be the
row labels with colname rowname: rownames can also be a
string giving the desired column name (see ‘Examples’). If
colnames is true, the column names are copied into row 1. This
is intended for cases where case conversion alters the original column
names and it is desired that they are retained. Note that there are
drawbacks to this approach: it presupposes that the rows will be
returned in the correct order; not always valid. It will also cause
numeric columns to be returned as factors.
Argument addPK = TRUE causes the row names to be marked as a
primary key. This is usually a good idea, and may allow database
updates to be done. However, the ODBC drivers for some DBMSs
(e.g. Access) do not support primary keys, and earlier versions of the
PostgreSQL ODBC driver generated internal memory corruption if this
option is used.
sqlUpdate updates the table where the rows already exist. Data
frame dat should contain columns with names that map to (some
of) the columns in the table. It also needs to contain the column(s)
specified by index which together identify the rows to be
updated. If index = NULL, the function tries to identify such
columns. First it looks for a primary key for the table, then for the
column(s) that the database regards as the optimal for defining a row
uniquely (these are returned by sqlColumns(special =
TRUE): if this returns a pseudo-column it cannot be used as we do not
have values for the rows to be changed). Finally, the row names are
used if they are stored as column "rownames" in the table.
When fast = TRUE, NAs are always written as SQL nulls in
the database, and this is also the case if fast = FALSE and
nastring = NULL (its default value). Otherwise nastring
gives the character string to be sent to the driver when NAs
are encountered: for all but the simplest applications it will be
better to prepare a data frame with non-null missing values already
substituted.
If fast = FALSE all data are sent as character strings.
If fast = TRUE, integer and double vectors are sent as types
SQL_C_SLONG and SQL_C_DOUBLE respectively. Some drivers
seem to require fast = FALSE to send other types,
e.g. datetime. SQLite's approach is to use the data to determine
how it is stored, and this does not work well with fast = TRUE.
If tablename contains . and neither catalog nor
schema is supplied, an attempt is made to interpret
qualifier.table names as table table
in schema qualifier (and for MySQL ‘schema’ means
‘database’). (This can be suppressed by opening the connection with
interpretDot = FALSE.)
Value
1 invisibly for success (and failures cause errors).
Warning
sqlSave(safer = FALSE) uses the ‘great white shark’
method of testing tables (bite it and see). The logic will
unceremoniously DROP the table and create it anew with its own
choice of column types in its attempt to find a writable
solution. test = TRUE will not necessarily predict this
behaviour. Attempting to write indexed columns or writing to
pseudo-columns are less obvious causes of failed writes followed by a
DROP. If your table structure is precious it is up to you back
it up.
Author(s)
Michael Lapsley and Brian Ripley
See Also
sqlFetch, sqlQuery,
odbcConnect, odbcGetInfo
Examples
## Not run:
channel <- odbcConnect("test")
sqlSave(channel, USArrests, rownames = "state", addPK=TRUE)
sqlFetch(channel, "USArrests", rownames = "state") # get the lot
foo <- cbind(state=row.names(USArrests), USArrests)[1:3, c(1,3)]
foo[1,2] <- 222
sqlUpdate(channel, foo, "USArrests")
sqlFetch(channel, "USArrests", rownames = "state", max = 5)
sqlDrop(channel, "USArrests")
close(channel)
## End(Not run)