You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The actual behavior currently differs between data types; for numerical types, the user-provided files are moved as explained above; for stings, the files are read and processed, i.e., the used-provided file remains in place as-is. We need to consider unifying this behavior, e.g., by deleting also the user provided files for strings once they have been processed/loaded.
In case the copy into file, e.g., due to a failing sanity check after the file is moved, that data is factually lost, i.e., the files are gone, but no data is loaded into the database; we need to either document this clearly, or make the implementation behave like a transaction.
I think we should actually reconsider the implementation (and hence also the documentation): we should probably always copy the data, and then the data also doesn't have to be on the same disk partition as the database.
For several years now, COPY BINARY copies the data from the source files. It does not in any way modify the referenced files.
I have also updated the documentation.
The text was updated successfully, but these errors were encountered:
Date: 2013-07-19 13:55:00 +0200
From: @yzchang
To: Documentation maintainers <>
Version: unspecified
CC: @mlkersten
Last updated: 2019-01-18 13:39:32 +0100
Comment 18925
Date: 2013-07-19 13:55:00 +0200
From: @yzchang
Triggered by thread "Must the file be deleted after "copy binary into table from file"" on July 19, 2013:
The actual behavior (move) is actually different than the copy that the documentation at http://www.monetdb.org/Documentation/Cookbooks/SQLrecipes/BinaryBulkLoad suggests. We need to fix the latter.
The actual behavior currently differs between data types; for numerical types, the user-provided files are moved as explained above; for stings, the files are read and processed, i.e., the used-provided file remains in place as-is. We need to consider unifying this behavior, e.g., by deleting also the user provided files for strings once they have been processed/loaded.
In case the copy into file, e.g., due to a failing sanity check after the file is moved, that data is factually lost, i.e., the files are gone, but no data is loaded into the database; we need to either document this clearly, or make the implementation behave like a transaction.
Comment 18928
Date: 2013-07-19 14:09:17 +0200
From: @mlkersten
Comment 21008
Date: 2015-07-15 17:09:14 +0200
From: @sjoerdmullender
I think we should actually reconsider the implementation (and hence also the documentation): we should probably always copy the data, and then the data also doesn't have to be on the same disk partition as the database.
Comment 26808
Date: 2019-01-18 13:39:32 +0100
From: @sjoerdmullender
For several years now, COPY BINARY copies the data from the source files. It does not in any way modify the referenced files.
I have also updated the documentation.
The text was updated successfully, but these errors were encountered: