Обсуждение: newline character handling
It is all postgresql all day, today :-) sorry if this should have gone to the novice forum. As I tried, using COPY, to import a few flat files created under Windows into postgresql running on a Linux machine, I discovered that: * If the last field in your record is a string, postgresql imports it, but keeps the ^M as part of the text string. * If the last field is numeric, postgresql refuses to import that line (because of the ^M, the field is not recognized as a number) Once I stripped the ^M, the data bulkloaded without a problem. Perhaps COPY should be smarter and recognize the DOS-style line endings? Curiously, krishna
> From: "Sampath, Krishna" <KSampath@ekmail.com> > To: pgsql-general@postgresql.org > Subject: newline character handling > Date: Fri, 7 Apr 2000 15:49:58 -0400 > > As I tried, using COPY, to import a few flat files created under Windows > into postgresql running on a Linux machine, I discovered that: > * If the last field in your record is a string, postgresql imports it, but > keeps the ^M as part of the text string. > * If the last field is numeric, postgresql refuses to import that line > (because of the ^M, the field is not recognized as a number) > > Once I stripped the ^M, the data bulkloaded without a problem. Perhaps COPY > should be smarter and recognize the DOS-style line endings? I'm ok with this for numerics, but against it for text. Why? Because I work with some binary data, and I wouldn't want the mysterious problem of not being able to COPY a line containing a record that's _supposed_ to end in ^M. -- Mike
maybe we need a keyword DOS|UNIX or perhaps TEXT|BINARY to tell postgresql to pick DOS style or UNIX style line endings... krishna -----Original Message----- From: Michael Blakeley [mailto:mike@blakeley.com] Sent: Saturday, April 08, 2000 12:57 PM To: pgsql-general@hub.org Subject: [GENERAL] Re: newline character handling > From: "Sampath, Krishna" <KSampath@ekmail.com> > To: pgsql-general@postgresql.org > Subject: newline character handling > Date: Fri, 7 Apr 2000 15:49:58 -0400 > > As I tried, using COPY, to import a few flat files created under Windows > into postgresql running on a Linux machine, I discovered that: > * If the last field in your record is a string, postgresql imports it, but > keeps the ^M as part of the text string. > * If the last field is numeric, postgresql refuses to import that line > (because of the ^M, the field is not recognized as a number) > > Once I stripped the ^M, the data bulkloaded without a problem. Perhaps COPY > should be smarter and recognize the DOS-style line endings? I'm ok with this for numerics, but against it for text. Why? Because I work with some binary data, and I wouldn't want the mysterious problem of not being able to COPY a line containing a record that's _supposed_ to end in ^M. -- Mike
> maybe we need a keyword DOS|UNIX or perhaps TEXT|BINARY to tell postgresql > to pick DOS style or UNIX style line endings... Maybe we just need to make sure that the files we are using are in the correct format for the platform they're being processed on. ; ) steve