Suggestion needed on data storage format in text file

Manish

The project I am developing doesn't involves database. I want to parse
the mailbox file (.mbx) and store the summary in the text file for fast
retrieval and display of information in the Inbox page.

The sugegsted format are as:

#1

ID [4 bytes]: Subject [100 bytes]: To Address[100 bytes]: From
Address[100 bytes]...etc...

#2

Instead of preassining fixed size to variable (as actual data may be
much less or can grew to more), we can store the values continuously,
seperated by some unique seperator (#|#, *#*, ...)

1324#|#Hi, How are yo******@google.com#|****@google.com#|# ... and so
on
Which of these will be the efficeint one (as there will be frequent
insert/delete/update of the individual information, eg. set message as
read ..., delete message ..., new message ...)

Also please suggest on how to determine the variable size (100 bytes as
in #1), and assign the size to the variable accordingly and read it
(differentiate multiple variables) when required.

Thanks.

Manish

Jul 19 '06 #1

Subscribe Post Reply

2352

Jerry Stuckle

Manish wrote:

The project I am developing doesn't involves database. I want to parse
the mailbox file (.mbx) and store the summary in the text file for fast
retrieval and display of information in the Inbox page.

The sugegsted format are as:

#1

ID [4 bytes]: Subject [100 bytes]: To Address[100 bytes]: From
Address[100 bytes]...etc...

#2

Instead of preassining fixed size to variable (as actual data may be
much less or can grew to more), we can store the values continuously,
seperated by some unique seperator (#|#, *#*, ...)

1324#|#Hi, How are yo******@google.com#|****@google.com#|# ... and so
on
Which of these will be the efficeint one (as there will be frequent
insert/delete/update of the individual information, eg. set message as
read ..., delete message ..., new message ...)

Also please suggest on how to determine the variable size (100 bytes as
in #1), and assign the size to the variable accordingly and read it
(differentiate multiple variables) when required.

Thanks.

Manish

Personally, I'd use a database. I wouldn't even try a flat file for
this. Too much work trying to keep things straight.

But you asked about the formats. The fixed length fields will have
extra space any time the amount of data is less than that of the amount
reserved. Then you run into the problem of someone who gets very
verbose with their subject line and exceeds the 100 characters. And 4
bytes allows up to 9999 ID's. Is that enough? Or are you going to try
to read/write binary (not easy in PHP)?

The second one is problematical because the user may include your
separator in its Subject: line (or even name/address if you pick the
wrong character).

Two other ways - use CSV format, which is well documented and supported
by PHP and other programs. Or, add a length field at the beginning of
each field, specifying how many characters in the following field.

But I'd still use a database.
--
==================
Remove the "x" from my email address
Jerry Stuckle
JDS Computer Training Corp.
js*******@attglobal.net
==================

Jul 19 '06 #2

Andy Jeffries

On Tue, 18 Jul 2006 21:28:13 -0700, Manish wrote:

#1
ID [4 bytes]: Subject [100 bytes]: To Address[100 bytes]: From Address[100
bytes]...etc...

#2
1324#|#Hi, How are yo******@google.com#|****@google.com#|# ... and so on
Which of these will be the efficeint one (as there will be frequent
insert/delete/update of the individual information, eg. set message as
read ..., delete message ..., new message ...)

The first one will be more efficient from a search/replace point of view,
the second will be more efficient from a space usage point of view.
Efficiency is subjective.

Also please suggest on how to determine the variable size (100 bytes as
in #1), and assign the size to the variable accordingly and read it
(differentiate multiple variables) when required.

substr would be used to cut out various portions of the string (e.g. 100
charactes starting at position 4) and sprintf (or fprintf to do it in PHP5
if you're using PHP5 to save a step).

If you need more than a pointer to the right functions, then it's starting
to sound like a homework assignment and I wish you luck with it...

Cheers,
Andy
--
Andy Jeffries MBCS CITP ZCE | gPHPEdit Lead Developer
http://www.gphpedit.org | PHP editor for Gnome 2
http://www.andyjeffries.co.uk | Personal site and photos

Jul 19 '06 #3

ronverdonk

4,258

Expert 4TB

Don't re-ivent the wheel!
Either use a data base or, if you must, stay with well documented formats like CSV or XML.

Ron :cool:

Jul 19 '06 #4

ronverdonk

4,258

Expert 4TB

Don't re-ivent the wheel!
Either use a data base or, if you must, stay with well documented formats like CSV or XML.

Ron :cool:

And to contradict myself:
I just saw a new class 'Variable Length Coding' at the PHP Classes, link:
http://www.phpclasses.org/browse/package/3232.html

Its description reads:
This class can be used to compress and uncompress data using the variable length encoding.

It can read a stream of data and pack it using an pure PHP implementation of the variable length encoding algorithm.

It can also do the opposite reading a variable length encoded stream of data and unpack it to restore the original uncompressed data.

So, if you still want variable length coding, check it out!

Ronald :cool:

Jul 19 '06 #5

Chung Leong

Manish wrote:

The project I am developing doesn't involves database. I want to parse
the mailbox file (.mbx) and store the summary in the text file for fast
retrieval and display of information in the Inbox page.

The sugegsted format are as:

#1

ID [4 bytes]: Subject [100 bytes]: To Address[100 bytes]: From
Address[100 bytes]...etc...

#2

Instead of preassining fixed size to variable (as actual data may be
much less or can grew to more), we can store the values continuously,
seperated by some unique seperator (#|#, *#*, ...)

1324#|#Hi, How are yo******@google.com#|****@google.com#|# ... and so
on
Which of these will be the efficeint one (as there will be frequent
insert/delete/update of the individual information, eg. set message as
read ..., delete message ..., new message ...)

Also please suggest on how to determine the variable size (100 bytes as
in #1), and assign the size to the variable accordingly and read it
(differentiate multiple variables) when required.

Thanks.

Manish

That's the kind of project that SQLite was designed for. It's worth
looking into.

Jul 19 '06 #6

ImOk

My suggestion is to use XML. PHP and Javascript has the Dom class that
supports this format very well. Its also easily extensible. And best of
all it's a text file.

Sample:

<mailbox name="some user">
<email>
<id>1234</id>
<subject>Send me the check<subject>
<to>no****@nospam.com</to>
<from>so*****@someone.com</from>
<message><![CDATA[blah blah blah blah blah
blah blah blah]]></message>
<attach>path to attach 1</attach>
<attach>path to attach 2</attach>
</email>
<email>
<id>5678</id>
<subject>Send me the check<subject>
<to>no****@nospam.com</to>
<from>so*****@someone.com</from>
<message><![cdata[blah blah blah ]]></message>
<attach>path to attach 1</attach>
<attach>path to attach 2</attach>
</email>
....etc...
</mailbox>
<mailbox name="some other user">
....
</mailbox>

Chung Leong wrote:

Manish wrote:
The project I am developing doesn't involves database. I want to parse
the mailbox file (.mbx) and store the summary in the text file for fast
retrieval and display of information in the Inbox page.

The sugegsted format are as:

#1

ID [4 bytes]: Subject [100 bytes]: To Address[100 bytes]: From
Address[100 bytes]...etc...

#2

Instead of preassining fixed size to variable (as actual data may be
much less or can grew to more), we can store the values continuously,
seperated by some unique seperator (#|#, *#*, ...)

1324#|#Hi, How are yo******@google.com#|****@google.com#|# ... and so
on
Which of these will be the efficeint one (as there will be frequent
insert/delete/update of the individual information, eg. set message as
read ..., delete message ..., new message ...)

Also please suggest on how to determine the variable size (100 bytes as
in #1), and assign the size to the variable accordingly and read it
(differentiate multiple variables) when required.

Thanks.

Manish

That's the kind of project that SQLite was designed for. It's worth
looking into.

Jul 20 '06 #7

Manish

Hi Jerry Stuckle, the project specifies not to use database, otherwise
it would have been definitely much easier. I have to store all the
information in the file itself. Thanks for bringing into atention that
whatever, seperator with least probbability of occurence is chosen, it
can occur in subject line. May be we should use some escape character
for it. As it is used in mailbox file. Every new mail starts with "From
", but if it's in the message itself, it's replaced by ">From ". I will
also look into the CSV format for storing the data.
Hi Andy Jeffries, we are using PHP 5, so sprintf/fprintf can be used. I
haven't come across using pointers in PHP. I will definitely try to
learn it.
Hi ImOk, our initial datastructure was in the XML format itself,
(individual XML file for every user). As there can be thousands of
email, the file will grew larger and reading/writing may be slow/error
prone. So it was suggested to use text file.

-----------------------------------------------------------------------------------------------------------------------------
This is how the datastructure is
-----------------------------------------------------------------------------------------------------------------------------

<mails>
<details id="">

<mailtype></mailtype>

<istemplate></istemplate>

<mailboxid></mailboxid>

<priority></priority>

<isstarred></isstarred>

<isread></isread>

<isreplied></isreplied>

<isforwarded></isforwarded>


<hasattachment></hasattachment>

<attachments>
<attdetails id="">

<filename></filename>

<filesize></filesize>
</attdetails>
</attachments>

<fromname></fromname>

<fromemail></fromemail>

<totalconversation></totalconversation>

<mainemailsno></mainemailsno>

<toemails></toemails>

<ccemails></ccemails>


<htmlcontent></htmlcontent>

<textcontent></textcontent>

<sentdatetime></sentdatetime>

<messagesize></messagesize>


<offsetinmbx></offsetinmbx>


<incomingdetails>
</incomingdetails>
<outgoingdetails>

<bccemails></bccemails>

<msgstatus></msgstatus>

<signatureid></signatureid>

<scheduledtime></scheduledtime>

<requestreceipt></requestreceipt>

<sendstatus></sendstatus>
</outgoingdetails>
</details>

</mails>

-----------------------------------------------------------------------------------------------------------------------------

But the other setting will still be in XML file.

We are using SimpleXML functions (get values, update values), DOM
(insert). Still the delete functionality is not working. We are
thinking of implementing preg_replace() for it.

Thanks.

Manish

Jul 20 '06 #8

Andy Jeffries

On Wed, 19 Jul 2006 21:07:06 -0700, Manish wrote:

Hi Andy Jeffries, we are using PHP 5, so sprintf/fprintf can be used. I
haven't come across using pointers in PHP. I will definitely try to learn
it.

It's not pointers but string parsing (getting out a section of a string
and formatting a string to contain exact lengths of string).

Hi ImOk, our initial datastructure was in the XML format itself,
(individual XML file for every user). As there can be thousands of email,
the file will grew larger and reading/writing may be slow/error prone. So
it was suggested to use text file.

I don't wish to sound offensive, but if you can't correctly write to an
XML file without errors, why do you think you'll be able to do it to a
flat file using functions/methods you don't know?

Also, bear in mind if you use a database it will also handle locking from
multiple processes easily, which you will have to handle yourself in this
situation.

Don't think "we'll only have one user accessing their account through a
single web instance so we won't have concurrency issues" - people these
days may use browser tabs to work on their mail concurrently.

And you really do run the risk of data loss/corruption if you don't
correctly lock access to the file.

Cheers,

Andy

--
Andy Jeffries MBCS CITP ZCE | gPHPEdit Lead Developer
http://www.gphpedit.org | PHP editor for Gnome 2
http://www.andyjeffries.co.uk | Personal site and photos

Jul 20 '06 #9

Jerry Stuckle

Manish wrote:

Hi Jerry Stuckle, the project specifies not to use database, otherwise
it would have been definitely much easier. I have to store all the
information in the file itself. Thanks for bringing into atention that
whatever, seperator with least probbability of occurence is chosen, it
can occur in subject line. May be we should use some escape character
for it. As it is used in mailbox file. Every new mail starts with "From
", but if it's in the message itself, it's replaced by ">From ". I will
also look into the CSV format for storing the data.
Hi Andy Jeffries, we are using PHP 5, so sprintf/fprintf can be used. I
haven't come across using pointers in PHP. I will definitely try to
learn it.
Hi ImOk, our initial datastructure was in the XML format itself,
(individual XML file for every user). As there can be thousands of
email, the file will grew larger and reading/writing may be slow/error
prone. So it was suggested to use text file.

-----------------------------------------------------------------------------------------------------------------------------
This is how the datastructure is
-----------------------------------------------------------------------------------------------------------------------------

<snip>

-----------------------------------------------------------------------------------------------------------------------------

But the other setting will still be in XML file.

We are using SimpleXML functions (get values, update values), DOM
(insert). Still the delete functionality is not working. We are
thinking of implementing preg_replace() for it.

Thanks.

Manish

Manish,

If the problem is speed, a flat file isn't going to help you that much
more. You'll still have to encode and decode the data, no matter which
format you use. And even if it's faster now, all you're doing is
delaying the inevitable. You definitely need a database.

If it were me, I'd go back to them and explain why they need a database.
But I'm only a consultant...

--
==================
Remove the "x" from my email address
Jerry Stuckle
JDS Computer Training Corp.
js*******@attglobal.net
==================

Jul 20 '06 #10

Chung Leong

ImOk wrote:

My suggestion is to use XML. PHP and Javascript has the Dom class that
supports this format very well. Its also easily extensible. And best of
all it's a text file.

XML and any text format is very inefficiency when updates/deletions are
frequent, as you have to rewrite the file everytime. For a mailbox,
that's unacceptable since the file size will likely be fairly large. A
suitable format requires a directory of sort storing the offsets of
records, so you can quicly seek to the them and modify them in place.
Whatever you come up with it'll end up resembling a database. So why
not just use what's there already?

Jul 20 '06 #11

ImOk

Agreed,

But I believe there are database engines whose natural format is XML..
It's probably fixed length.

Chung Leong wrote:

ImOk wrote:
My suggestion is to use XML. PHP and Javascript has the Dom class that
supports this format very well. Its also easily extensible. And best of
all it's a text file.

XML and any text format is very inefficiency when updates/deletions are
frequent, as you have to rewrite the file everytime. For a mailbox,
that's unacceptable since the file size will likely be fairly large. A
suitable format requires a directory of sort storing the offsets of
records, so you can quicly seek to the them and modify them in place.
Whatever you come up with it'll end up resembling a database. So why
not just use what's there already?

Jul 21 '06 #12

Manish

>I don't wish to sound offensive, but if you can't correctly write to an
XML file without errors, why do you think you'll be able to do it to a
flat file using functions/methods you don't know?

>Also, bear in mind if you use a database it will also handle locking from
multiple processes easily, which you will have to handle yourself in this situation.

>Don't think "we'll only have one user accessing their account through a
single web instance so we won't have concurrency issues" - people these
days may use browser tabs to work on their mail concurrently.

>And you really do run the risk of data loss/corruption if you don't
correctly lock access to the file.

It's definitely a serious issue. Opening same files concurrently, for
each tabbed browser and then update the content of index file will be
less effecient.

e.g. There can be >1000 messages, say 2 are unread, user reads 1
message, to update the status from read to unread for that message, we
have to update single byte positin for that message. It's critical from
performance (response to user) point of view. If we do it in database,
it will be much faster.

>If the problem is speed, a flat file isn't going to help you that much more. You'll
still have to encode and decode the data, no matter which format you use. And
even if it's faster now, all you're doing is delaying the inevitable. You definitely
need a database.

>If it were me, I'd go back to them and explain why they need a database.
But I'm only a consultant...

Surely. We will also suggest for the database.

>XML and any text format is very inefficiency when updates/deletions are
frequent, as you have to rewrite the file everytime. For a mailbox,
that's unacceptable since the file size will likely be fairly large. A
suitable format requires a directory of sort storing the offsets of
records, so you can quicly seek to the them and modify them in place.

The mailbox file (.mbx) will be there. We will parse it and store only
some of the details (including mailbox file offset for that message) in
the index file. (.idx, .xml, and surely the best will be database)

Jul 21 '06 #13

Chung Leong

Manish wrote:

Surely. We will also suggest for the database.

Keep in mind that using a "database" doesn't necessarily imply a
full-blown, standalone RDBMS. An embedded database like SQLite or
Sleepycat would work very well in these types of situations.

Jul 21 '06 #14

Andy Jeffries

On Thu, 20 Jul 2006 13:58:59 -0700, Chung Leong wrote:

XML and any text format is very inefficiency when up
dates/deletions are frequent, as you have to rewrite the file
everytime.

Actually, that's not strictly true. If your text format file has fixed
length records, you could do random writing to it if you fopen it using
"a+" access mode and then use fseek/fwrite to overwrite just the record
you're interested in. No need to rewrite the whole file every time.

Cheers,
Andy

--
Andy Jeffries MBCS CITP ZCE | gPHPEdit Lead Developer
http://www.gphpedit.org | PHP editor for Gnome 2
http://www.andyjeffries.co.uk | Personal site and photos

Jul 25 '06 #15

Suggestion needed on data storage format in text file

Similar topics