Skip to content

Latest commit

 

History

History
263 lines (218 loc) · 6.76 KB

README.md

File metadata and controls

263 lines (218 loc) · 6.76 KB
Raw

An Extentsion Package for DBI. Implementation of new generic function dbUpdateTable

Installation

devtools::install_github("schardtbc/DBIExt")

Introduction

S4 generic dbUpdateTable()

This DBI Extention package introduces a new S4 generic dbUpdateTable. Updating tables is not the focus of the DBI package as pure analysis scripts treat a database tables as read-only.

In other domains, database maintainence requires updating information in rows of an existing table. Dropping the table and recreating it as is done by dbWriteTable with the overwrite argument set to TRUE is not an optimal solution for many applications.

No single generic update function could cover the diverse variety of possible UPDATE queries. The dbUpdateTable S4 generic introduced in this package performs a specific type of DBMS UPDATE operation using the update join syntax. An example of this syntax is shown in the sql fragment shown below. This syntax is used by MariaDB and MySQL.

UPDATE T1
INNER JOIN T2 ON T1.C1 = T2.C1 [AND T1.C2 = T2.C2]
SET T1.CX = T2.CX,
    T1.CY = T2.CY;

reference: tutorial on UPDATE join syntax for MySQL

Note that no WHERE clause is specified for the query. This by design; the reasoning for which will be explained later in the document.

For dbUpdateTable() the table T1 must be a pre-existing table in the DBMS. Table T2 will be created as a temporary table in the DBMS using a dataset supplied as a function argument. This dataset will usually be a tibble, a data.frame, or another object coercible to a data.frame.

the dbUpdateTable function call has two additional arguements set and by used for constructing the SET and ON clauses of the UPDATE query.

dbUpdateTable(con, "T1", T2, set = c("CX","CY"), by = c("C1","C2"))

SQL Helper Functions

There is a helper function sqlUpdateTable which generates the UPDATE sql but does not execute it. Helper functions are also provided to generate the sql for the ON clause of the join as well as the multiples SET clauses of the UPDATE. These are sqlOnClause() and sqlSetClause().

T2 <- tibble::as_tibble(list(C1=1:26,C2=LETTERS,CX=0,CY=1))
sqlUpdateTable(con, "T1", T2, set = c("CX","CY"), by = c("C1","C2"))
sqlOnClause(con,"T1","T2",by = c("C1","C2"));
sqlSetClause(con,"T1","T2",set = c("CX","CY"));

Table Indexing

It is best practice that the target DBMS table be indexed, with the indexed columns matching the columns specified by the by argument. The temporary table used store the update dataset is indexed by the columns in the by argument as part of the dbUpdateTable() implementation. The UPDATE-JOIN operation with indexed table is generally quite fast.

Why no WHERE caluse?

Update queries can use a WHERE clause to limit the scope of the update. However in an UPDATE-JOIN the scope of the query is alreadly limited by the ON clause of the join. This combined with the power of R to precisely form the update dataset makes use of a WHERE clause redundent, IMHO.

A (very) Simple Example

Connection to local test database

library(DBI)
library(DBIExt)

config <- list(
    HOST = "localhost",
    DATABASE = "mktdata000",
    USER = "test",
    PASSWORD = "test");

#' get the main database connection
getDBConnection <- function(){
  connection <- DBI::dbConnect(RMariaDB::MariaDB(), host = config[["HOST"]],
                               user = config[["USER"]], password = config[["PASSWORD"]], dbname = config[["DATABASE"]])
}

#' close the database connection
closeDBConnection <- function(connection) {
  DBI::dbDisconnect(connection)
}

db <- getDBConnection();
db
#> <MariaDBConnection>
#>   Host:    localhost
#>   Server:  8.0.13
#>   Client:  8.0.13

Setup a DBMS table

cid <- 1:26;
cn  <- 100:125;
clc  <- letters[cid];


target.df <- tibble::as_tibble(list(id = cid, cn = cn, cc = clc));
DBI::dbWriteTable(db,"T123",target.df,overwrite = TRUE);

original <- dbReadTable(db, "T123")
original
#>    id  cn cc
#> 1   1 100  a
#> 2   2 101  b
#> 3   3 102  c
#> 4   4 103  d
#> 5   5 104  e
#> 6   6 105  f
#> 7   7 106  g
#> 8   8 107  h
#> 9   9 108  i
#> 10 10 109  j
#> 11 11 110  k
#> 12 12 111  l
#> 13 13 112  m
#> 14 14 113  n
#> 15 15 114  o
#> 16 16 115  p
#> 17 17 116  q
#> 18 18 117  r
#> 19 19 118  s
#> 20 20 119  t
#> 21 21 120  u
#> 22 22 121  v
#> 23 23 122  w
#> 24 24 123  x
#> 25 25 124  y
#> 26 26 125  z

Setup a data.frame for the update data

cuc  <- LETTERS[cid];
update.df <- tibble::as_tibble(list(id = cid, cc = cuc));
update.df <- dplyr::group_by(update.df,id)

update.df
#> # A tibble: 26 x 2
#> # Groups:   id [26]
#>       id cc   
#>    <int> <chr>
#>  1     1 A    
#>  2     2 B    
#>  3     3 C    
#>  4     4 D    
#>  5     5 E    
#>  6     6 F    
#>  7     7 G    
#>  8     8 H    
#>  9     9 I    
#> 10    10 J    
#> # … with 16 more rows

Perform the Update In The DBMS

res <- dbUpdateTable(db,"T123",update.df);
print(glue::glue("number of records modified in DBMS {res}"))
#> number of records modified in DBMS 26

updated <- dbReadTable(db,"T123");
closeDBConnection(db);

updated
#>    id  cn cc
#> 1   1 100  A
#> 2   2 101  B
#> 3   3 102  C
#> 4   4 103  D
#> 5   5 104  E
#> 6   6 105  F
#> 7   7 106  G
#> 8   8 107  H
#> 9   9 108  I
#> 10 10 109  J
#> 11 11 110  K
#> 12 12 111  L
#> 13 13 112  M
#> 14 14 113  N
#> 15 15 114  O
#> 16 16 115  P
#> 17 17 116  Q
#> 18 18 117  R
#> 19 19 118  S
#> 20 20 119  T
#> 21 21 120  U
#> 22 22 121  V
#> 23 23 122  W
#> 24 24 123  X
#> 25 25 124  Y
#> 26 26 125  Z

Setting Indexes on Existing Tables

dbCreateTable(conn = db,
              name = "daily_time_series",
              fields <- list(
                symbol = "VARCHAR(8)",
                date   = "DATE",
                sourceName = "VARCHAR(4)",
                sequenceID = "INTEGER",
                datetime = "DATETIME",
                epoch = "DOUBLE",
                open = "FLOAT",
                high = "FLOAT",
                low = "FLOAT",
                close = "FLOAT",
                volume = "DOUBLE",
                aOpen = "FLOAT",
                aHigh = "FLOAT",
                aLow = "FLOAT",
                aClose = "FLOAT",
                aVolume = "DOUBLE",
                dividendAmount = "FLOAT",
                splitCoefficient = "FLOAT"
              );

# add the composit primary key                            
dbAddPrimaryKey(conn = db, name = "daily_time_series", primary_key = c("symbol","date","sourceName")) 

# add seconary imdexes on date and sequenceID
dbAddIndex(conn = db, name = "daily_time_series", index = "sequenceID")
dbAddIndex(conn = db, name = "daily_time_series", index = "date")