Options
creating a new table in databases
crappy_viking
Member Posts: 16 
Maven
Hello All,
While working with RapidMiner, I realized that there is a really tedious work to do when modifying attributes' names and types. Explaining myself :
- I work on a database, on a specific table where I work either in readonly or in RW mode. I mean, I have to read or write attributes and values on a given table. Typically on a postgres or mysql table.
- This kind of problem is not that important when you have only few attributes, but with feature selection and high dimensionality context (text mining ? others ?) you have to manipulate hundreds of attributes and types.
- The idea is to read attributes from the database, and directly create an attributeConstruction object/file from what has been just read.
- On the other hand, if you have an attributeConstruction object/file, you should just try to create a new table according to this construction (hundreds of attributes), using a simple SQL client, and I hope you will agree it is a very tedious job.
- Thus one would need two new operators : one for reading a table's metadata and creating an attributes' construction, another for writing in the database, creating "if not exists" a new table whose metadata fit the AttributeConstruction supplied at its input.
Do you agree ? Would it be feasible ?
Cheers,
CP.
While working with RapidMiner, I realized that there is a really tedious work to do when modifying attributes' names and types. Explaining myself :
- I work on a database, on a specific table where I work either in readonly or in RW mode. I mean, I have to read or write attributes and values on a given table. Typically on a postgres or mysql table.
- This kind of problem is not that important when you have only few attributes, but with feature selection and high dimensionality context (text mining ? others ?) you have to manipulate hundreds of attributes and types.
- The idea is to read attributes from the database, and directly create an attributeConstruction object/file from what has been just read.
- On the other hand, if you have an attributeConstruction object/file, you should just try to create a new table according to this construction (hundreds of attributes), using a simple SQL client, and I hope you will agree it is a very tedious job.
- Thus one would need two new operators : one for reading a table's metadata and creating an attributes' construction, another for writing in the database, creating "if not exists" a new table whose metadata fit the AttributeConstruction supplied at its input.
Do you agree ? Would it be feasible ?
Cheers,
CP.
0
