Remote Profiling with D7 (NOT memory profiling, but timing...) - delphi

We have one machine in-house that is 20 times slower starting our Delphi 7 app than any other machine.
We would like to get a performance profile (not memory profile) to locate where it's spending its time.
AQTime, which we own, we've discovered doesn't do remote profiling.
We'd prefer not to take the time to build up an entire D7 IDE development environment just so we can use AQTime to profile our app on this one in-house machine.
The code is a bit too complex for us to want to meter it ourselves.
Any suggestions on a profiler that will gather high level (procedure or line number) statistics remotely?

Take a look at SamplingProfiler. It doesn't do "remote" profiling, but it also doesn't require a development environment. It just needs to be able to launch the program to be profiled (so it has to run on the same machine) and the program has to have a .MAP file generated by the linker in the same folder as the .EXE. If this is in-house, that shouldn't be a problem for you.
And if you look at the helpfile, you'll even find ways to have it only profile certain sections of your program, which AQTime can't do. That helps if you know the issue is in one specific place, such as the startup code.

Why not install AQTime on the machine and use it as a standalone profiler? No need for an "entire D7 IDE development environment".

You can also try my free/open source sampling profiler:
http://code.google.com/p/asmprofiler/wiki/AsmProfilerSamplingMode
(I get better results with it than with SamplingProfiler)
It uses all kinds of Delphi debug symbols (.map, TD32, .jdbg, etc)

You can use our Open Source TSynLog class to add profiling to any application, not only on the developer computer.
It is not an automated profiler, as other tools: you'll have to modify your code. But it can be run on request remotely and even with no communication at all, even from the end customer side.
You add some profiling calls to some method code, then entering and leaving the methods will be logged into a text file. Then a supplied log viewer is available, and has some dedicated method to do the profiling, and identify the slow methods.
(source: synopse.info)
The logging mechanism can be used to trace recursive calls. It can use an interface-based mechanism to log when you enter and leave any method:
procedure TMyDB.SQLExecute(const SQL: RawUTF8);
var ILog: ISynLog;
begin
ILog := TSynLogDB.Enter(self,'SQLExecute');
// do some stuff
ILog.Log(sllInfo,'SQL=%',[SQL]);
end; // when you leave the method, it will write the corresponding event to the log
It will be logged as such:
20110325 19325801 + MyDBUnit.TMyDB(004E11F4).SQLExecute
20110325 19325801 info SQL=SELECT * FROM Table;
20110325 19325801 - MyDBUnit.TMyDB(004E11F4).SQLExecute 00.000.507
Here the method name is set in the code ('SQLExecute'). But if you have an associated .map file, the logging mechanism is able to read this symbol information, and write the exact line number of the event. You can even use a highly compressed version of the .map file (900 KB .map -> 70 KB .mab, i.e. much better than zip or lzma), or embed its content to the executable at build time.
Adding profiling at method level is therefore just the matter of adding one line of code at the beginning of the method, as such:
procedure TMyDB.SQLExecute(const SQL: RawUTF8);
begin
TSynLogDB.Enter;
// do some stuff
end; // when you leave the method, it will write the corresponding event to the log
High-resolution timestamps are also logged on the file (here 00.000.507). With this, you'll be able to profile your application with data coming from the customer side, on its real computer. Via the Enter method (and its auto-Leave feature), you have all information needed for this.
By procedding steps by steps, you'll get very quickly to your application bottlenecks. And it would be possible to do the same on the end customer side, on request.
I used this on several applications, and found out very easily several bottlenecks, even on specific hardware, software and network configuration (you never know what your customers use), very easily.

Related

AV after successful close of applications

I am getting this AV message about 3 to 5 seconds after the applications close as expected:
Exception EAccessViolation in module rtl160.bpl at 00073225. Access violation at address 500A3225 in module 'rtl160.bpl'. Read of address 00000004.
These (20) applications are very similar in that they are IBX business applications. About half of them did not cause the AV to occur.
These applications were ported from Delphi-xe and they worked flawlessly for a long time. No changes were made to the projects in the port. Both 32 and 64 bit builds gave the same results.
Is this a bug in some library's finalization section freeing a resource or something?
I am using Delphi-XE2 Update 3.
Would appreciate the help.
Try using madExcept / EurekaLog etc. - they give you detailed stack trace on AV. This is not always a panacea, but can point you to the problem.
Access Violations are by their nature already very troublesome beasts since they deal with invalid pointers in memory. One that occurs a while after an application shuts down is even worse because that's when your app is in "cleanup" mode. You're could be dealing with something that went wrong much earlier in the application, but is only exposing itself at shutdown.
General Tips:
Try to always undo things in the reverse order you did them. E.g.
Create A, Create B ... Destroy B, Destroy A
Connect to Database, Open Dataset ... Close Dataset, Disconnect from Database
Even making sure you've done all the above before shutting down can help tremendously.
Any threads that are still running while your application is running can cause problems.
Preferably ensure all your child threads are properly terminated before final shutdown.
Refer back to Closing datasets above. Depending on what you're doing, some database components will create their own threads.
If you're using COM, try ensure ComObj is high up in the initialization sequence (I.e. place it as high as possible in your DPR).
Delphi finalizes units in the reverse order that they were initialized.
And you don't want ComObj to finalize before other things that are dependent on ComObj have also done so.
If you're using interface references, make sure you resolve circular reference issues.
Some of these problems can be tricky to find, but you can do the following:
Setup a source-code "sandbox" environment (you're going to chuck all your changes as soon as you've found the problem).
Figure out the simplest set of steps required to guarantee the error. (Start app and immediately shutdown would be ideal.)
Then you're going to comment-out delete wipe out chunks of code between tests and basically follow a divide and conquer approach to:
rip out code
test
if the problem persists, repeat. Else roll-back and rip out a different chunk of code.
eventually your code base will be small enough to pinpoint likely problems which can be tackled with targeted testing.
I've had this kind of access violation problem on occasion with old Delphi or C++Builder projects. Today I had it with C++Builder. At the time of the crash, by looking in the Debug -> Call Stack window, I can see that it's happening inside a call to fflush, called by __exit_streams and _exit.
I'm not sure what is causing it, since it's so deep in the Borland library code, but it seems to come and go at random when the code changes. And it seems to be more common with multi-form applications.
This time the error went away when I just added a new button on the main form. A button which is just there, has no event handlers and does not do anything. I think that any random change to the code, classes, variables etc rearranges the memory layout when you relink the application, and that either triggers or untriggers the error.
For now, I just leave the new button on the form, set it to "not visible" so that there's no visible change. As it seems to work, it's good enough solution for me at this time.

How can I keep a large amount of OutputDebugString() calls from degrading my application in the Delphi 6 IDE?

This has happened to me on more than one occasion and has led to many lost hours chasing a ghost. As typical, when I am debugging some really difficult timing-related code I start adding tons of OutputDebugString() calls, so I can get a good picture of the sequence of related operations. The problem is, the Delphi 6 IDE seems to be able to only handle that situation for so long. I'll use a concrete example I just went through to avoid generalities (as much as possible).
I spent several days debugging my inter-thread semaphore locking code along with my DirectShow timestamp calculation code that was causing some deeply frustrating problems. After having eliminated every bug I could think of, I still was having a problem with Skype, which my application sends audio to.
After about 10 seconds the delay between my talking and hearing my voice come out of Skype on the second PC that I was using for testing, the far end of the call, started to grow. At around 20 - 30 seconds the delay started to grow exponentially and at that point triggered code I have that checks to see if a critical section was being held too long.
Fortunately it wasn't too late at night and having been through this before, I decided to stop relentlessly tracing and turned off the majority of the OutputDebugString(). Thankfully I had most of them wrapped in a conditional compiler define so it was easy to do. The instant I did this the problems went away, and it turned out my code was working fine.
So it looks like the Delphi 6 IDE starts to really bog down when the amount of OutputDebugstring() traffic is above some threshold. Perhaps it's just the task of adding strings to the Event Log debugger pane, which holds all the OutputDebugString() reports. I don't know, but I have seen similar problems in my applications when a TMemo or similar control starts to contain too many strings.
What have those of you out there done to prevent this? Is there a way of clearing the Event Log via some method call or at least a way of limiting its size? Also, what techniques do you use via conditional defines, IDE plug-ins, or whatever, to cope with this situation?
A similar problem happened to me before with Delphi 2007. Disable event viewing in the IDE and instead use DebugView from Sysinternals.
I hardly ever use OutputDebugString. I find it hard to analyze the output in the IDE and it takes extra effort to keep several sets of multiple runs.
I really prefer a good logging component suite (CodeSite, SmartInspect) and usually log to various files. Standard files for example are "General", "Debug" (standard debug info that I want to collect from a client installation as well), "Configuration", "Services", "Clients". These are all set up to "overflow" to a set of numbered files, which allows you to keep the logs of several runs by simply allowing more numbered files. Comparing log info from different runs becomes a whole lot easier that way.
In the situation you describe I would add debug statements that log to a separate logfile. For example "Trace". The code to make "Trace" available is between conditional defines. That makes turning it on pretty simple.
To avoid leaving in these extra debug statements, I tend to make the changes to turn on the "Trace" log without checking it out from source control. That way, the compiler of the build server will throw out "identifier not defined" errors on any statements unintentionally left in. If I want to keep these extra statements I either change them to go to the "Debug" log, or put them between conditional defines.
The first thing I would do is make certain that the problem is what you think it is. It has been a long time since I've used Delphi, so I'm not sure about the IDE limitations, but I'm a bit skeptical that the event log will start bogging down exponentially over time with the same number of debug strings being written in a period of 20-30 seconds. It seems more likely that the number of debug strings being written is increasing over time for some reason, which could indicate a bug in your application control flow that is just not as obvious with the logging disabled.
To be sure I would try writing a simple application that just runs in a loop writing out debug strings in chunks of 100 or so, and start recording the time it takes for each chunk, and see if the time starts to increase as significantly over a 20-30 second timespan.
If you do verify that this is the problem - or even if it's not - then I would recommend using some type of logging library instead. OutputDebugString really loses it's effectiveness when you use it for massive log dumps like that. Even if you do find a way to reset or limit the output window, you'd be losing all of that logging data.
IDE Fix Pack has an optimisation to improve performance of OutputDebugString
The IDE’s Debug Log View also got an optimization. The debugger now
updates the Log View only when the IDE is idle. This allows the IDE to
stay responsive when hundreds of OutputDebugString messages or other
debug messages are written to the Debug Log View.
Note that this only runs on Delphi 2007 and above.

Delphi logging with multiple sinks and delayed classification?

Imagine i want to parse a binary blob of data. If all comes okay, then all the logs are INFO, and user by default does not even see them. If there is an error, then user is presented with error and can view the log to see exact reason (i don't like programs that just say "file is invaid. for some reason. you do not want to know it" )
Probably most log libraries are aimed at quickly loading, classifying and keeping many many log lines per second. which by itself is questionable, as there is no comfort lazy evaluation and closures in Delphi. Envy Scala :-)
However that need every line be pre-сlassified.
Imagine this hypothetical flow:
Got object FOO [ok]
1.1. found property BAR [ok]
1.1.1. parsed data for BAR [ok]
1.2 found property BAZ [ok]
1.2.1 successfully parsed data for BAR [ok]
1.2.2 matching data: checked dependancy between BAR and BAZ [fail]
...
So, what can be desired features?
1) Nested logging (indenting, subordination) is desired then.
Something like highlighted in TraceTool - see TraceNode.Send Method at http://www.codeproject.com/KB/trace/tracetool.aspx#premain0
2) The 1, 1.1, 1.1.1, 1.2, 1.2.1 lines are sent as they happen in a info sink (TMemo, OutputDebugString, EventLog and so one), so user can see and report at least which steps are complete before error.
3) 1, 1.2, 1.2.2 are retroactively marked as error (or warning, or whatever) inheriting from most specific line. Obviously, warning superseeds info, error superseeds warning and info, etc/
4) 1 + 1.2 + 1.2.2 can be easily combined like with LogMessage('1.2.2').FullText to be shown to user or converted to Exception, to carry the full story to human.
4.1) Optionally, with relevant setup, it would not only be converted to Exception, but the latter even would be auto-raised. This probably would require some kind of context with supplied exception class or supplied exception constructing callback.
5) Multisink: info can be just appended into collapsible panel with TMemo on main form or currently active form. The error state could open such panel additionally or prompt user to do it. At the same time some file or network server could for example receive warning and error grade messages and not receive info grade ones.
6) extra associated data might be nice too. Say, if to render it with TreeView rather than TMemo, then it could have "1.1.1. parsed data for BAR [ok]" item, with mouse tooltip like "Foo's dimensions are told to be 2x4x3.2 metres"
Being free library is nice, especially free with sources. Sometimes track and fix the bug relying solely on DCUs is much harder.
Non-requiring extra executable. it could offer extra more advanced viewer, but should not be required for just any functionality.
Not being stalled/abandoned.
ability to work and show at least something before GUI is initialized would be nice too. Class constructors are cool, yet are executed as part of unit visualization, when VCL is not booted yet. If any assertion/exception is thrown from there, user would only see Runtime error 217, with all the detail lost. At least OutputDebugStreen can be used, if nothing more...
Stack tracing is not required, if needed i can do it and add with Jedi CodeLib. But it is rarely needed.
External configuration is not required. It might be good for big application to reconfigure on the fly, but to me simplicity is much more important and configuration in code, by calling constructors or such, is what really matters. Extra XML file, like for Log4J, would only make things more fragile and complex.
I glanced few mentioned here libraries.
TraceTool has a great presentation, link is above. Yet it has no info grade, only 3 predefined grades (Debug/Error/Warning) and nothing more, but maybe Debug would suit for Info replacement... Seems like black box, only saving data into its own file, and using external tool to view it, not giving stream of events back to me. But their messages nesting and call chaining seems cool. Cools is also attaching objects/collection to messages.
Log4D and Log4Delphi seems to be in a stasis, with last releases of 2007 and 2009, last targeted version Delphi 7. Lack documentation (probably okay for log4j guy, but not for me :- ) Log4Delphi even had test folder - but those test do not compile in Delphi XE2-Upd1. Pity: In another thread here Log4delphi been hailed for how simple is to create custom log appender (sink)...
BTW, the very fact that the only LOG4J was forked into two independent Delphi ports leaves the question of which is better and that both lack something, if they had to remain in split.
mORMot part is hardly separated from the rest library. Demo application required UAC escalation for use its embedded SQLite3 engine and is frozen (no window opened, yet the process never exits normally) if refused Admin grants. Another demo just started infinite stream of AV exceptions, trying to unwind the stack. So is probably not ready yet for last Delphi. Though its list of message grades is excessive, maybe even a bit too many.
Thank you.
mORMot is stable, even with latest XE2 version of Delphi.
What you tried starting were the regression tests. Among its 6,000,000 tests, it includes the HTTP/1.1 Client-Server part of the ORM. Without the Admin rights, the http.sys Server is not able to register the URI, so you got errors. Which makes perfectly sense. It's a Vista/Seven restriction, not a mORMot restriction.
The logging part can be used completely separated from the ORM part. Logging is implemented in SynCommons.pas (and SynLZ.pas for the fast compression algorithm used for archival and .map embedding). I use the TSynLog class without any problem to log existing applications (even Delphi 5 and Delphi 6 applications), existing for years. The SQLite3 / ORM classes are implemented in other units.
It supports the nesting of events, with auto-leave feature, just as you expect. That is you can write:
procedure TMyClass.MyMethod(const Params: integer);
begin
TSynLog.Enter;
// .... my method code
end;
And adding this TSynLog.Enter will be logged with indentation corresponding to the recursive level. IMHO this may meet your requirements. It will declare an ISynLog interface on the stack, which will be freed by Delphi at the "end;" code line, so it will implement an Auto-Leave feature. And the exact unit name, method name and source code line number will be written into the log (as MyUnit.TMyClass.MyMethod (123)), if you generated a .map file at compilation (which may be compressed and appended to the .exe so that your customers logs will contain the source line numbers). You have methods at the ISynLog interface level to add some custom logging, including parameters and custom state (you can log objects properties as JSON if you need to, or write your custom logging data).
The exact timing of each methods are tracked, so you are able to profile your application from the data supplied by your customer.
If you think the logs are too much verbose, you have several levels of logging, to be customized on the client side. See the blog articles and the corresponding part of the framework documentation (in the SynCommons part). You've for instance "Fail" events and some custom kind of events. And it is totally VCL-independent, so you can use it without GUI or before any GUI is started.
You have at hand a log viewer, which allow client-side profiling and nested Enter/Leave view (if you click on the "Leave" line, you'll go back to the corresponding "Enter", e.g.):
If this log viewer is not enough, you have its source code to make it fulfill your requirements, and all the needed classes to parse and process the .log file on your own, if you wish. Logs are textual by default, but can be compressed into binary on request, to save disk space (the log viewer is able to read those compressed binary files). Stack tracing and exception interception are both implemented, and can be activated on request.
You could easily add a numeration like "1.2.1" to the logs, if you wish to. You've got the whole source code of the logging unit. Feel free to ask any question in our forum.
Log4D supports nested diagnostic contexts in the TLogNDC class, they can be used to group together all steps which are related to one compound action (instead of a 'session' based grouping of log events). Multi-Sinks are called Appenders in log4d and log4delphi so you could write a TLogMemoAppender with around twentyfive lines of code, and use it at the same time as a ODSAppender, a RollingFileAppender, or a SocketAppender, configurable at run time (no external config file required).

Delphi: 30-day trial

How can I make a 30-day trial for my application? I need to allow users to use an application only 30 days. How to count these days?
I keep the first and the last date in registry. But if to change a system time - no protection will be. I need to count these 30 days.
You could probably come up with a system that requires an internet connection, but without something that the user can't tamper with, I don't see a solution.
Any solutions that rely on an untrusted element (an element of the protection that is under the user's control) is critically weak.
The simplest way I can think of to protect against the user moving the clock back is to limit the total number of launches.
However, attempts to limit the number of launches requires persistence -- saving data to the disk, perhaps encrypting and storing a modified version of your activation data file -
Imagine that you count one of the 30 days as "used up" once the app has been launched, on a unique occasion, even when the same date is re-used. In order to avoid using up more than 1 "activation time day" when launched, the user must allow your software to re-save its activation file each time it runs.
To block that approach, the user needs only to keep the apparent date from changing, plus they must either prevent you from storing anything to disk; or they can simply track and record your changes and reverse them out, either using a monitoring process, or using VMWare snapshots. About VMWare snapshots, you can do nothing. The virtual machine's disk is not under your control.
You can protect your app of users setting the clock back simply by storing in the registry the date of last execution.
Each time the app is started you need to do the following:
Check current date (as reported by the system clock) against the stored last execution date and, if current date is earlier than the last execution one, consider that the trial period has expired (or whatever you prefer).
If the previous check is ok, save the current date in the registry and continue execution.
As WarrenP says, any technique storing information locally can be easily circumvented using VMware snapshots.
And anyone, including those who check via internet, can be skipped via assembler level hacking.
Here's a discussion on Shareware trial enforcement with Delphi:
Best Shareware lock for Delphi Win32
Along with discussions on various 3rd-party solutions, techniques for DIY, etc..
IMO, DIY is feasible if your app produces data that the user will want to keep around, then you can simply embed a copy of the usage/day counter in the database in such a way that they can't wipe it without destroying their data. I also like watermarking (printing "trial" on reports, etc..), escalating nag severity, but I do not recommend or condone "drop-dead" crippling until WAY past the expiration data. I also like to measure "days of actual use", instead of using a calendar.
Registry manipulation works, and many of the 3rd-party protectors use it. But you need to be stealthy, and keep backups in several locations simultaneously.
You should also consider having separate trial and registered versions. But also consider that pirates will buy the registered version with a stolen card, and put it on Rapidshare, BitTorrent, etc..
Also note that elaborate defenses lead to support headaches. Sometimes PCs crash and the clock gets set backwards. They install new harware. PCs get rebuilt, restored from backup, etc.. If a user is running a debugger, he may be a software developer, not a pirate. If your app looks like it has been patched, it may be an overly-aggressive antivirus. And at any time, a shoddy patch for Windows may cause your program to think that it's being attacked, hacked, or reverse-engineered. You have been warned...
Encrypt a date and store it in registry the best way to do it is that date to be stored by the installer itself and if the date doesn't exist the application should quit.
There is an open source project (which was a commercial product before):
TurboPower OnGuard is a library to create demo versions of your Borland Delphi & C++Builder applications. Create demo versions that are time-limited, feature-limited, limited to a certain number of uses, or limited to a certain # of concurrent network users.
I have not checked which Delphi versions are supported.
For this kind of "protection" and some others, I have used TmxProtector (open source) from MaxComponents in the past with good results. From the link provided:
The TmxProtector is a software protection component. It was designed
for quick implementation of application protection functions. You can
create time-trial and password protected applications. You can set the
maximum number of execution, and it can work with registration keys as
well.
This compoment uses very simple encryption to store the expiration date in the registry and it provides some simple detection for tampering on the system date.
It sounds like you need to store the date the last registry entry was written. Then inside your program, test if the current date is less than the date last registry entry was made. If true display a message that the trial period has expired and the program must be purchased.
Here are some ideas on how to deal with clock changes during the trial period:
Save both the date of first and the date of the last program start. If the date of the last program start is greater than the current date, then the user has moved the clock back. I simply increase a day and save the new date as the date of last start. You can of course decide to just end the trial.
To try to defeat trial bypass programs (RunAsDate for example) which run your application by setting the date and time to a specific value, you can instead of getting the date via the usual Delphi way (Date, Now), get, for example, the last modification date of NTUSER.DAT.
Save your trial data on two separate locations, either two registry locations, or file and registry. This way even if the user deletes one of the trial data locations, you'll still have a backup one to use.
If you keep your trial info in registry, the registry could be deleted by the user. Evey one expects to find the registration info there.
There is one place where the user might not think to look into: your own app (EXE file). Put an ANSI string constant (MUST by ansi/ascii or other 1 byte string, static array, etc) into your program, like 'xyxyxyxyxyxy'. Compile your app. Open your complied app with a hex editor. Search for that string. Now your program could use that area to store the trial info into itself.
Use this method in conjunction with others: store your info in registry also, on disk, etc.
Anyway, the best would be to get the registration info from your server.
The big drawback: 1. The server must be ALWAYS online! 2. The user must be connected to internet (when it uses your app).
Also use a Delphi license management library to help you encrypt the license info and generate a string-based key that you can send to your customers upon registration.
Anyway, whatever you send to your server needs to be based on the hardware fingerprint of that computer. Otherwise your license key will leak out on some warez website and everyone will be able to use that key. But if the key is hardware-based it would be useless if it is leaked on Internet.
Just remember: don't over do it! There is no such thing as unbreakable software protection. Microsoft could not do it!
As the thread pointed to mentioned, I encourage you to look into WinLicense: http://www.oreans.com.
I've been using it for quite some time and it handles trial periods quite well. It also handles licensing, customer lists, etc.
Tom

Delphi Search files and directories fastest alghorithm

I'm using Delphi7 and i need a solution to a big problem.Can someone provide me a faster way for searching through files and folders than using findnext and findfirst? because i also process the data for each file/folder (creation date/author/size/etc) and it takes a lot of time...I've searched a lot under WinApi but probably I haven't see the best function in order to accomplish this. All the examples which I've found made in Delphi are using findfirst and findnext...
Also, I don't want to buy components or use some free ones...
Thanks in advance!
I think any component that you'd buy, would also use findfirst/findnext. Recursively, of course. I don't think there's a way to look at every directory and file, without actually looking at every directory and file.
As a benchmark to see if your code is reasonably fast, compare performance against WinDirStat http://windirstat.info/ (Just to the point where it's gathered data, and is ready to build its graph of the space usage.)
Source code is available, if you want to see what they're doing. It's C, but I expect it's using the same API calls.
The one big thing you can do to really increase your performance is parse the MFT directly, if your volumes are NTFS. By doing this, you can enumerate files very, very quickly -- we're talking at least an order of magnitude faster. If all the metadata you need is part of the MFT record, your searches will complete much faster. Even if you have to do more reads for extra metadata, you'll be able to build up a list of candidate files very quickly.
The downside is that you'll have to parse the MFT yourself: There's no WinAPI functions for doing it that I'm aware of. You also get to worry about things that the shell normally does for you in worrying about things like hardlinks, junctions, reparse points, symlinks, shell links, etc.
However, if you want speed, the increase in complexity is the only way to achieve it.
I'm not aware of any available Delphi code that already implements an MFT parser, so you'll probably have to either use a 3rd party library or implement it yourself. I was going to suggest the Open Source (GPL) NTFS Undelete, which was written in Delphi, but it implements the MFT parsing via Python code and has a Delphi-Python bridge built in.
If you want to get really fast search results consider using the Windows Search (API) or the Indexing service.
Other improvements might be to make use of threads and split the search for files and the gathering of file properties or just do a threaded search.
I once ran into a very similar problem where the number of files in the directory, coupled with findfirst/findnext was taking more time than was reasonable. With a few files its not an issue, but as you scale upwards into the thousands, or tens of thousands of files, then performance drops considerably.
Our solution was to use a queue file in a separate directory. As files are "added" to the system they were written to a queue file (was a fixed record file). When the system needed to process data, it would see if the file existed, and if so then rename it and open the renamed version (this way adds could occur for the next process pass). The file was then processed in order. We then archived the queue file & processed files into a subdirectory based on the date and time (for example: G:\PROCESSED\2010\06\25\1400 contained the files run at 2:00 pm on 6/25/2010).
Using this approach not only did we reach an almost "real-time" processing of files (delayed only by the frequency by which we processed the queue file), but we also insured processing of files in the order they were added.
If you need to scan remote drive with that many files, I would strongly suggest doing so with a "client-server" design, so that the actual file scanning is always done locally and only the results are fetched remotely. That would save you a lot of time. Also, all "server" could scan in parallel.
If your program is running on Windows 7 or Server 2008 R2, there are some enhancements to the Windows FindFirstFileEx function which will make it run a bit faster. You would have to copy and modify the VCL functions to incorporate the new options.
There isn't much room for optimization with a findfirst / findnext loop, because it's mostly I/O bound: the operating system needs to read this information from your HDD!
The proof: Make a small program that implements a simple findfirst / findnext loop that does NOTHING with the files it finds. Restart your computer and run it over your big directory, note the time it takes to finish. Then run it again, without restarting the computer. You'll notice the second run is significantly faster, because the operating system cached the information!
If you know for sure the directory you're trying to scan is heavily accessed by the OS because of some other application that's using the data (this would put the directory structure information into the OS's cache and make scanning not be bound to the I/O) you can try running several findfirst/findnext loops in parallel using threads. The down side of this is that if the directory structure is not allready in the OS cache your algorithm is again bound to HDD in/out and it might be worst then the original because you're now making multiple parallel I/O requests that need to be handled by the same device.
When I had to tackle this same problem I decided against parallel loops because the SECOND run of the application is allways so much faster, prooving I'm bound to I/O and no ammount of CPU optimisation would fix the I/O bottleneck.
I solved a similar problem by using two threads. This way I could "process" the file(s) at the same time as they where scanned from the disk. In my case the processing was significantly slower than scanning so I also had to limit the number of files in memory at one time.
TMyScanThread
Scan the file structure, for each "hit" add the path+file to a TList/TStringList or similar using Syncronize(). Remember to Sleep() inside the loop to let the OS have some time too.
PseudoCode for the thread:
TMyScanThread=class(TThread)
private
fCount : Cardinal;
fLastFile : String;
procedure GetListCount;
procedure AddToList;
public
FileList : TStringList;
procedure Execute; Override;
end;
procedure TMyScanThread.GetListCount;
begin
fCount := FileList.Count;
end;
procedure TMyScanThread.AddToList;
begin
FileList.Add(fLastFile);
end;
procedure TMyScanThread.Execute;
begin
try
{ Get the list size }
Syncronize( GetListCount );
if fCount<500 then
begin
// FindFirst code goes here
{ Add a file to the list }
fLastFile := SR.Name; // Store Filename in local var
Syncronize( AddToList ); // Call method to add to list
SleepEx(0,True);
end else
SleepEx(1000,True);
finally
Terminate;
end;
end;
TMyProcessFilesThread
Get the oldest entry in the list, and Process it. Then output results to DB.
This class is implemented similarly with Syncronized methods that access the list.
One alternative to the Syncronize() calls is to use a TCriticalSection. Implementing Syncronization between threads is often a matter of taste and the task at hand ...
You can also try BFS vs. DFS. This may affect your performance.
Link
http://en.wikipedia.org/wiki/Breadth-first_search
http://en.wikipedia.org/wiki/Depth-first_search
When I started to run into performance problems working with lots of small files on in the file system I moved to storing the files as blobs in database. There is no reason why related information like size, creation, and author couldn't also be stored in the database. Once the tables are populated in the database, I suspect that the database engine could do a much faster job of finding records (files) than any solution that we are going to come up with since Database code is highly specialized for efficient searches through large data sets. This will definitely be more flexible since adding a new search would be as simple as creating a new Select statement. Example: Select * from files where author = 'bob' and size > 10000
I'm not sure that approach will help you. Could you tell us more about what you are doing with these files and the search criteria.

Resources