moodymudskipper / easydb Goto Github PK

I don't want to support partial matching but partial matching is associated with $.

If column not found it would be nice to check if partial matching would have worked, and if so, have the error inform the user that partial matchng doesn't work and that if he means "this table" he should ask for it explicitly.

I think we shouldn't build methods for class "DBIConnection". Instead db_connect() should create a class "easydb" and we'll have "easydb" methods.

quite convenient :

summary.DBIConnection <- function(object, ...){
  cat(crayon::cyan("OBJECT"), "\n",sep="")
  print(object)
  cat("\n", crayon::cyan("TABLES"), "\n",sep="")
  print(DBI::dbListTables(object))
  cat("\n","use `str(object)` for details", "\n", sep="")
  
  cat("\n", crayon::cyan("DEFAULT summary"), "\n",sep="")
  summary.default(object,...)
}

str.DBIConnection <- function(object, ...){
  cat(crayon::cyan("OBJECT"), "\n",sep="")
  print(object)
  cat("\n", crayon::cyan("TABLES"), "\n",sep="")
  tables <- DBI::dbListTables(object)
  tables <- setNames(tables, tables)
  for(tbl in tables){
    cat("\n", crayon::yellow(tbl), "\n", sep="")
    error <- try(tibble::glimpse(object[[tbl]]),silent = TRUE)
    if(inherits(error,"try-error")) cat(error,"\n")
  }
  
  cat("\n", crayon::cyan("DEFAULT str"), "\n",sep="")
  invisible(utils:::str.default(object,...))
}

if we add a class to our easydb_tbl class to our tbl_lazy objects we can have a [.easydb_tbl method.

This can act a bit like data table, and := will work to "update where". But we skip all the sophisticated stuff, grouping etc...

Maybe would be good to have this here if not implemented in tibble. easydb already imports tibble.

view <- function(x, title = NULL, ...){
  UseMethod("view")
}

view.default <- tibble::view()

view.tbl_lazy <- function(x, title = NULL, n = 500, ...){
  if (is.null(title)) title <- rlang::expr_deparse(substitute(x)) # enexpr doesn't work in methods, see https://github.com/r-lib/rlang/issues/368
  View(dplyr::collect(head(x, n)), title, ...)
}

Maybe this could be used for connections instead of
#3

Reprex below.

library(easydb)
library(dplyr)
#> 
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag
#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union

## Try 1
drv <- DBI::dbDriver("Oracle")
con <- db_connect(
  drv,
  dbname = "my_db",
  username = "read",
  password = "read",
  schema = "my_schema"
)

con$my_tbl
#> Error in .oci.SendQuery(conn, statement, data = data, prefetch = prefetch,  : 
#>  ORA-00942: table or view does not exist

## Try 2
specs <- list(
  drv = DBI::dbDriver("Oracle"),
  dbname = "my_db",
  username = "read",
  password = "read"
)

con <- db_connect(specs)

my_tbl <- dbplyr::in_schema("my_schema", "my_tbl")

con$my_tbl
#> Error in .oci.SendQuery(conn, statement, data = data, prefetch = prefetch,  : 
#>  ORA-00942: table or view does not exist

# works
dplyr::tbl(con, my_tbl)

^{Created on 2019-09-02 by the reprex package (v0.2.1)}

• list tables in DB
• list symbols in expr : https://stackoverflow.com/questions/57670823/how-can-i-list-all-symbols-used-in-a-call
• find intersection
• substitute foo by con$foo when relevant
• eval

it is wrong at the moment, in particular for PostgreSQL ?RPostgres::Postgres is not useful at all and user should do ?RPostgres::dbConnect,PqDriver-method``

It would be nice to have direct links to those too.

because we used this : getFromNamespace("db_has_table.DBIConnection", "dbplyr")

I'm not sure why we used a non exported function there but there should be an exported alternative