Corey Nelson wrote:

>Are you guys sure?
>

I think we have a concensus here, yeah :-D

>
>Just kidding.  ;)  Thanks guys.
>
>What about just the reading part? Could one read page 935 without loading
>the whole file into memory? That doesn't seem like it should be hard to
>do... just read some sort of index and skip to the part you need. Surely
>they got that much right.
>
>Corey
>

Take a look at the event-driven APIs for POIFS and HSSF -- yes, I'm sure 
that much can be done. Bear in mind, I'm not personally involved in the 
HDF implementation at the moment, and I'm assuming we can create a 
similar API ... I could be wrong about that.

You have a high regard for the company that invented these file formats 
... perhaps you need your medication levels adjusted? :-D

>
>
>>-----Original Message-----
>>From: Marc [SMTP:[EMAIL PROTECTED]]
>>Sent: Monday, March 04, 2002 7:15 PM
>>To:   [EMAIL PROTECTED]
>>Subject:      Re: Fast loading and saving?
>>
>>acoliver wrote:
>>
>>>>On Mon, 4 Mar 2002 15:48:00 -0700 Corey Nelson
>>>>
>><[EMAIL PROTECTED]>
>>
>>>wrote.
>>>
>>>>I've been told that one of the benefits (perhaps the only) of using OLE2
>>>>
>>CD
>>
>>>>format is that you don't have to load or save the entire file when you
>>>>
>>just
>>
>>>>want read or edit a small part of it. If you had War and Peace in a Word
>>>>Document and you wanted to fix a typo on page 935 it wouldn't take
>>>>
>>forever
>>
>>>>to open the file or save your changes.
>>>>
>>>>Question 1) Is this true?
>>>>
>>>NO!  Thats totally bogus!  OLE 2 CDF is formatted as such to make this
>>>impossible (downstream pointers to upstream things and vice versa).
>>>
>>>Marc, would you like to expand on that?
>>>
>>Ahem. Sure, Andy.
>>
>>Basically, OLE 2 CDF is similar to an old fashioned file system (how 
>>similar? similar enough that it would benefit from a defragmenter tool 
>>...). The only way for an application to perform the efficient editing 
>>you speak of, it would need to directly manipulate the file system. 
>>Imagine an application that took over the writing of individual sectors 
>>"for efficiency sake" ... that's what an application would have to do to 
>>achieve the efficiency you speak of -- it would have to manipulate the 
>>block chains to do that.
>>
>>My experience with very large Word documents (~50-60 pages) indicates 
>>that this efficiency does not exist. Said documents are slow to load and 
>>slow to write.
>>
>>>
>>>>Question 2) If so, is this still true if I'm using POIFS to read and
>>>>
>>write
>>
>>>>the file?
>>>>
>>>Nope, we read it rather efficiently if I might say so, but the format
>>>
>>SUCKS
>>
>>>and prevents this.  
>>>
>>We read it with as much efficiency as we can bring to bear, given the 
>>inefficient file format. We write it out as efficiently as possible, 
>>more efficiently than other *cough* applications seem to do (our output 
>>requires no defragmentation).
>>
>>>
>>><rant>
>>>
>><snip/>
>>
>>></rant>
>>>
>>>OLE 2 Compound Document format is NOTHING more then a completely
>>>
>>convoluted
>>
>>>archive format.  Use ZIP with no compression and you achieve the same
>>>thing..   The only thing OLE 2 CDF buys you is that it slighly obfuscates
>>>(poorly) the format of your file from platforms without a Java 2 JVM
>>>
>>(since
>>
>>>POI unobfuscates it ;-) )...  
>>>
>>>So why would we put ourselves through this:  Its fun!  Intellectually
>>>challenging!  (It shouldn't be to just read a file format, but it IS!)
>>>
>>And
>>
>>>we get to spit in the eye of closed file formats everywhere!
>>>
>>>>Thanks for reading.
>>>>
>>>No problem..  I needed to unwind for the day ;-)
>>>
>>>-Andy
>>>
>>>PS to understand what I think of the OLE 2 Compound Document Format, XLS
>>>format or DOC format --  count all of the times the word SUCKS appears in
>>>this email and multiply it by 10 to the 99th power and you will have some
>>>vague idea on what I think of the formats and the morons who designed
>>>
>>them.
>>
>>>:-)  -- But unmangling them into nice object models is really fun! 
>>>(seriously we do this for the intellectual challenge because all of the
>>>
>>work
>>
>>>people pay us for is to technically easy!)
>>>
>>>
>>>>Corey
>>>>
>>>
>>
>



Reply via email to