Forum OpenACS Development: cr_write_content and file return...

Collapse
1: cr_write_content and file return...
Posted by Jeff Davis on 10/21/02 05:30 PM
I noticed that thumbnails were not being cached from photo-album and tracked it down to the call in cr_write_content which calls ns_cpfp to return the filesystem file. If I change cr_write_content to look like 
switch $storage_type {
        text { 
            ReturnHeaders $mime_type
            ns_write [db_string write_text_content ""]
        }
        file {
            set path [cr_fs_path $storage_area_key]
            set filename [db_string write_file_content ""]
            ns_returnfile 200 $mime_type $filename
        }
        lob  {
            ReturnHeaders $mime_type
            db_write_blob write_lob_content ""
        }
    }
and I now get 304s for unchanged files. This is not 100% safe since the revision file timestamp is not necessarily the timestamp you should use for the item but it seems like we should make cr_write_content pay attention to if-modified-since headers and I think it should be pretty straightforward to do so. Does anyone have any objection?
Collapse
2: Response to cr_write_content and file return... (response to 1)
Posted by Jun Yamog on 10/22/02 04:56 AM
Hi Jeff,

No objection its better than what we currently have.  But I believe someone brought this up in the bboards a few weeks ago.  And he/she also made some changes.  It would be nice if that someone will post here what he/she did.

Collapse
3: Response to cr_write_content and file return... (response to 1)
Posted by Don Baccus on 10/22/02 06:34 PM
I have no objection.  I wrote this routine but all I did was to bundle up code from the file storage package into a convenient utility (every package returning a file duplicated this functionality at the time).
Collapse
4: Response to cr_write_content and file return... (response to 1)
Posted by Peter Alberer on 10/24/02 11:47 PM
I posted the same issue a few weeks ago and supposed to change the code in db_write_blob. But doing the ns_returnfile immediately in cr_write_content is probably the better solution. Anyway i think its good to do this change as like Jeff mentioned, attention should be payed to if-modified-since headers.
Collapse
5: Response to cr_write_content and file return... (response to 1)
Posted by Jamie Rasmussen on 10/25/02 12:45 PM
There are related problems in db_exec_lob and cr_create_content_file.  Several of the file upload problems people have had with Win32 OpenACS occur because file handles aren't fconfigure'd with "-translation binary" before use.

See JCD's comment in cr_create_content_file - TCL's file copy really would be better than ns_cpfp.  Peter's changes to use ns_returnfile in db_exec_lob should fix the other problems - I've just been adding the fconfigure where necessary, but the proposed solution increases efficiency too.