Related
Lets say I'm making an app where I need to push updates to the client-side regularly. Now most of these update would only ever affect my own code, never the libraries that form up my dependancies. As far as I'm aware, the transformer I'll use when invoking pub build, will take all my libraries, dependancies, and whatever else, and compile them together in a single web/main.dart.js file. This tends to remind me somewhat of static linking, as with languages like C++.
For obvious reasons, compiling like this is something you'd only want to do when finally deploying an app. For most people, it suffices to just use dartium and work off the .dart hierarchy directly. However, what if I were testing in a javascript browser, trying out dart:js code, for instance. I wouldn't want to have to recompile all of angular and friends, when they go entirely untouched. My specific case is the desire to use CDN services for static files within my app.
AngularDart, as my example, contributes a massive 22,000 lines to my compiled javascript, and if I change one little thing within my own app, I can kiss 304 NOT MODIFIED goodbye, let alone CDN savings, for that bulk of untouched angular.
All that being said, is there a way to de-couple dependancies in a transformed dart application? Can I "dynamicly link" my dart libraries? And furthermore, could I hypothetically distribute my dart libraries as dart.js, for use in other transformed code?
No, it doesn't work this way. This could work if the deployable contained the entire dependencies but that's not the case.
pub build includes only the code that's actually used in your app (tree shaking) and this is usually only a small fraction of a dependency. Changing a single character in your code can require including different code from your dependencies. I think treeshaking has a much bigger effect than caching could have.
I have a large application to manage consisting of of three or four executables and as many as fifty .dlls. Many of the source code files are shared across many of the projects.
The problem is a familiar one to many of us - if I change some source code I want to be able to identify which of the binaries will change and, therefore, what it is appropriate to retest.
A simple approach would be simply to compare file sizes. That is an 80% acceptable solution, but there is at least a theoretical possibility of missing something. Secondly, it gives me very little indication as to WHAT has changed; It would be ideal to get some form of report on this so I can then filter out irrelevant (e.g. dates/versions copyrights etc..)
On the plus side :
all my .dcus are in a row - I mean they are all built into a single folder
the build is controlled by a script (.bat)(easy, for example, to emit .obj files if that helps)
svn makes it easy to collect together any (two) revisions for comparison
On the minus side
There is no policy to include all used units in all projects; some units get included because they are on a search path.
Just knowing that a changed unit is used/compiled by a project is not sufficient proof that the binary is affected.
Before I begin writing some code to solve the problem I would like to ask the panel what suggestions they might have as to how to approach this.
The rules of StackOverflow forbid me to ask for recommended software, but if anyone has any positive experiences of continuous integration tools that would help - great
I am open to any suggestion or observation that is relevant in this context.
It seems to me that your question boils down to knowing which units are contained in your various executables. Since you are using search paths, it will be hard for you to work this out ahead of time. The most robust way to find out is to consult the .map file that the compiler emits. This contains a list of all units contained in your executable.
Once you know which units are contained in each executable, you need to know whether or not anything has changed in those units. That information is contained in your revision control system. Put this all together and you have the information that you need.
Of course, just because the source code for a unit has changed, you might argue that re-testing is not needed. Perhaps the only change made was the version, or the date in a copyright label or some such. But it is asking too much to be able to ask a computer to make such a judgement. At some point you need a human to step up and take responsibility.
What is odd about this though is that you are asking the question at all. It seems to me to be enormously risky to attempt partial testing. I cannot understand why you don't simply retest the entire product.
After using it for > 10 years for commercial in-house and freelancer work in large projects, I can recommend to try Apache Ant. It is a build tool which supports dependencies, and has many very helpful features.
Apache Ant also integrates nicely with CI tools such as Hudson/Jenkins, Bamboo etc.
Another suggestion - based on experience with Maven - is to design the general software architecture as modular as possible. If modules (single or multiple source or DCU files in one directory) use a version number in the directory name as a version number, it is possible to control exactly how application are composed from these modules.
If you want to program such a tool yourself the approach would be something like this:
First you need to detect wheter there were any changes made to seperate source files. As you already figured out comparing the file size is bad idea as the file size can stay the same despite lots of changes made to it (as long as there is same amount of text in pas file its size won't change). So instead you could check the last modification time for specific file or create some hash value like MD5 hash for comparison (can be quite slow).
Then you need to generate yourself a dependancy tree which will tell you which files are used for which project/subproject.
Finally based on changes detected in seperate files you check the dependancy tree to see which projects needs to be recompiled.
The problem of such approach is that you would probably have to update the dependancy tree manually each time when new unit is added to the project or an existing one is removed from the project.
But the best way would be to go and use some version controll software istead of reinventing the wheel. I myself like the way how GIT works and I belive that with proper implementation of GIT into the project mannager itself could be quite powerfull do to GIT support of branching/subbranching (each project is its own branch, each version of your software can be its own subbranch).
Now latest version of Delphi does have GIT integration done though SVN but this unfortunately limits some of best GIT functionality. So if you maybe decide to go and integrate GIT support directly into Delphi I'm first in line to use it.
I want to open libraries, because currently I want to see the algorithms used for drawing, modify them and implement them in my program. For example: I have tried to create an algorithm on my own for lines. But I failed. And even if I had succeeded, I fear that it might not give the same result as the algorithm in the libraries. And I don't want this to happen. That's why I want to copy the algorithms used for the methods in libraries. And I really hope that this will help me create the application I'm currently working on and with other applications in the future.
I tried to open the libraries with a code editor. But I had troubles finding the libraries- I don't really know where are they placed nor in what files are their codes stored.
How to open a Java library? Or is there a place in the Internet where the code is uploaded?
It sounds like what you want is to get inside the standard Java libraries (so you can see the code for methods like Graphics.drawLine()).
You can download the source files from the same place you got the JDK, if you are on Windows or Linux. For the Mac, see this question. You can even set up Eclipse so that you can debug into that source as if it were your own code.
However, you will probably not find line-drawing code in Java in these libraries - the Graphics implementation will almost certainly use native methods, and may just call existing methods in the OS.
If you are specifically looking for line drawing algorithms, another option would be to look at the Wikipedia page for the Bresenham (aliased) or Wu (antialiased) algorithm.
Edit:
The part of a Graphics2D call that actually puts pixels on the screen is probably inside a system call and therefore the source would not be available.
A java vector graphics library like Batik might have source for some of these algorithms, but probably relies on the Graphics2D calls for most of them. So, you might look for a comprehensive vector graphics library written in a language other than Java, where those graphics calls do not already exist by default.
Alternately, checking the table of contents for a computer graphics book might point you at a variety of algorithms that you could look up on Wikipedia.
For any given library:
Make sure to obey all licenses when using another's code
If you are referring to the Java SDK source code, you can find it here: http://grepcode.com/
If the project is open source, you can usually just get the source from the project website. No problem, though make sure to obey their license.
If the project is NOT open source, well, then you're in a pickle licensing wise, so I do NOT endorse this, however, you would need to use a Java Decompiler such as JD-Gui
As far as what drawing algorithms to use, there are so many different ones (obviously, people have been trying to draw quickly for many many years), your best bet is to figure out exactly what you need to do and then search for that specific need separately. There isn't really a good repository of ALL of them, except maybe wikipedia.
If you are using the libraries they are on your classpath. Check out how to figure out your classpath in whichever IDE you are using and you can find the JARs you depend on. If they are packaged with sources all you need to do it unjar them and look at the sources.
If you don't have access to the sources you can get the code using a Java Decompiler.
If you are trying to look at a standard Java library, see the other answers about getting the source to the JDK.
If you are interested in an open source library (such as something maintained by the Apache project), look on the site of the project for a 'source jar' which you can open with a standard zip utility.
If the library you want is not open source or you cannot find the source for it, you can try to decompile it. If you are using Eclipse, try this decompiler.
I'm working on a large delphi 6 project with quite a lot of dependancies. It takes several minutes to compile the whole project. The recompilation after a few changes is sometimes much more longer so that it is quicker to terminate Delphi, erase all dcu files and recompile everything.
Does anyone know a way to identify, what makes the compiler slower and slower? Any tips how to organize the code to improve compiler performance?
I have already tried following things:
Explicitly include most of the units in the dpr instead of relying on the search path: It didn't improve anything.
Use the command line compiler dcc32: it isn't faster.
Try to see what the compiler does (using ProcessExplorer from SysInternals): apparently it runs most of the time a function called 'KibitzGetOverloads'. But I can't do anything with this information...
EDIT, Summary of the answers until now:
The answer that worked best in my case:
The function "Clean unused units references" from cnpack. It almost automatically cleaned more than 1000 references, making a "cold" compilation about twice faster. ("cold" compilation = erase all dcu files before compiling). It gets the reference list from the compiler. So if you have some {$IFDEF } check that all your configurations still compile.
The next thing I would like to try:
Refactoring the unit references manually (eventually using an abstract class)
but it is much more work, since I first need to identify where the problems are. Some tools that might help:
GExperts adds a project dependencies browser to the delphi IDE (but unfortunately it can not show the size of each branch)
Delphi Unit Dependency Viewer V1.0 do about the same thing but without Delphi. It can calculate some simple statistics (Which units is the most referenced, ...)
Icarus which is referenced on a link in one of the answer.
Things that didn't change anything in my case:
Putting every files from my program and all components in one folder without subfolders.
Defragmenting the disk (I tried with a ramdisk)
Using a ramdisk for the code source and output folders.
Turning off the live scanning antivirus
Listing all the units in the dpr file instead of relying on the search path.
Using the command line compiler dcc32 or ecc32.
Things that didn't apply to my case:
Avoiding having dependencies on network shares.
Using DelphiSpeedUp, because I already had it.
Using a single folder for all dcu (I always do it)
Things that I didn't try:
Upgrading to another Delphi version.
Using dcc32speed.exe
Using a solid-state drive (I didn't tried it, but I tried with a ramdisk where I put all the source code. But maybe I should have installed delphi on the ramdisk too)
Some things that could slow down the compiler
Redundant units in your uses clause. See this question for a link to CnPack.
Not explicitly adding units to your project file. You've already seem to have covered that.
Changed compiler settings, most notably include TDD32 info.
Try to get rid of unused units in your uses clause and see if it makes a difference.
using Delphi 7 and 2009, last week I pass from almost 2 minutes for compiling and another 45 seconds from hitting f9 and get the main form of my app to 20 seconds compiling and running. This things has drive me crazy for about 6 months and nothing I tried seems to work. Using filemon from SysInternals, I realize than every unit (mostly components) that compiler requires was searched in every folder that was in Search Path, yes, this produce a LOT of FileOpen, FileExists and FileNotFound, etc. What I do was, put every DCU, DFM, RES, etc from components all in a single folder, and having just this folder in the search path, and a couple of others folders required by the project; the results were amazing. Other problem prior to the fix, was debugging. It takes almos 40 seconds in each F7, F8 key press while debuging, this has been fixed too. Hope this info can help you. Greetings form Isla de Margarita, Venezuela. Excuse my english, if any error ;)
Check are there any paths in search paths that aren't on your local machine.
i.e. Don't link to binaries on network shares, and check that the search path isn't checking any network shares.
I haven't seen the compiler get slower over time, but it's been a long time since we used Delphi 6.
It seems to be generally agreed upon in the Delphi community that, if you don't want to upgrade to the latest and greatest (Delphi 2007 or 2009), then Delphi 7 is the best/fastest/most stable. You might consider upgrading.
KibitzGetOverloads sounds like something from the kibitz compiler -- the "background" compiler that gives you code-completion, background error highlighting, code tooltips, etc. Sounds like you'd be better off checking the call stack of the command-line compiler, not the IDE; you'd get something more helpful.
I have never found compiles to be faster after deleting the DCUs. DCUs are there to make the build incremental, therefore faster. If you're seeing faster compiles after deleting all DCUs, check your hardware. Have you defragged your hard disk lately? How much free space do you have on the drive?
Have you set a single folder to get the DCUs. If not, they will be scattered all over.
Put all the units and their implicitly called units (except installed components from Library path) in the dpr. To be sure you did not miss some, empty your search path, it still should compile.
After reducing the search path, you can try to reduce your library path by installing your components into fewer folders.
Although only partly relevant to your exact question, I hear that the use of a solid-state drive is vastly increasing compile time with Delphi - Nick Hodges said this himself on the Delphi Podcast a couple of week ago.
Brian
U can automatically get rid of
unnecesseary unit references, which is very efficient optimization for compiling speed.
In your situation, dividing your
project into packages can improve
compiling speed. With this way, it
just generates modified package(s),
not single massive binary for each
recompilation. Working with packages
can also help about easy deployment
of your project updates.
Turn off your live scanning antivirus
We had the same (or similar) problem.
I of our package has compilation Time about 12 min.
After changes, now we have moved to 32 sg.
After many tests we found that the "problematic situation" was the following:
In a single package:
The A unit uses a large number of units: U1, U2, U3, U4, ... U100 (Uses of Interface) in the same package. This is an important unit that centralizes all the initialization work.
All units of the package, U1, U2, U3, .., U100 uses unit A (use of implementation)
This "circular reference" does not give compilation errors because the USES are different, but caused a large compile-time.
SOLUTION:
Eliminate the reference to each unit, U1, U2, U3 ,...., U100 in the A Unit.
Now, A unit use a large number of units: U1, U2 ,...., U100, but the units U1, U2 ,..., U100, does not use the unit A.
After this change the compile-time is down drastically.
If you have a similar situation, you can try this.
Excuse for my bad english.
Greetings.
Neftalí -Germán Estévez-
I had the same problem and I can come up with (2) reasons it effected me.
Circular references. The gentleman who stated that one was correct. I would have certain LARGE projects that would compile fast, and SMALL projects that compiled slow. Could not figure it out until I restructured the code and then I got the faster compile speeds. Lots of small units. It's easy to build monolithic units. But, there are many penalties from it.
I've heard it a 1000 times, develop on a slow machine like your users might be using. Hey, that's for the testing department. I can't waste time with compiling, Delphi load speeds, packages, etc. I went out and bought a "GAMERS" computer (WOW) with the Solid State Drives (as mentioned earlier), 12GB RAM, OVERCLOCKED "i7" Intel chip, triple video cards (linked), all on Vista64 (Vista is not bad once it is finally running with all installed parts). It was a real pain to get it all set up. But, I am not waiting anymore on my computer. Pure compile speed, load speed, plus a new fresh machine without all of the crap that was installed on the last one over the last 2 years. I even unloaded DelphiSpeedUp. Did not need it. And I don't need to turn off AntiVirus, since I did that one as well, and got penalized with the internet crap. So AntiVirus stays on. Pure and simple, get a BALLS OUT machine. Your time is worth more than what you will spend on a new computer.
Try to install a ram disk and set your dcu output path to point there. This more than halved my compilation time with Delphi 2007 on top of DelphiSpeedUp.
The compiler will only compile units that have changed. If you have changed the code in the interface section all units that depend on the changed unit are compiled. If only code in the implementation section is changed, the compile will only that unit but presumably link all the modules. Implies a good design of interfaces up front but if you restructure the code to restrict changes to the implementation compile times might reduce. I have no idea by how much. This fact is mentioned in the Delphi help files under Multiple and indirect unit references in Delphi 7 "Using Delphi".
Do not compile on network drives. Seek time is dramatically worse.
Consider pointing your dcu ("unit output" directory to a ramdrive.
Limit the number of include/unit directories.
Try to avoid minor circular references that the compiler still accepts, specially for large units (e.g. generated ORM units for your OPF). It might cause large units to be compiled twice. (does Delphi allow minor mutual circulars, or is that a FPC only feature?)
I never tried, but hardcoding all files with full/relative path in the central .dpr might also help (script to regenerate/update?). (you mention that above, but was it with path xx in '\path\yyy' notation?).
Other long shots:
Use Kylix (file/dir I/O under Linux is dramatically better in my experience (though that is from FPC experience)). Maybe we need a reversed cross-kylix :-)
Use a separate (windows) build machine, and tweak NTFS over the registry to be less "safe". (which you don't care for, since everything is a revision system to begin with). Afaik these options can only be done global for all filesystems, hence the separate system. Throw in a raid array or Raptor too.
Forget solid state. Nice buzz atm, but the high write ratio will kill it eventually (both life and performance when it gets fuller and can't optimally allocate anymore), and you need the expensive intel ones to beat two $75 HD's in RAID.
P.s. Sorry for the FPC references. I do both, and I sometimes don't know anymore what belongs to what.
What I do is always make sure to have very few directories in the library path, and most of the components and static code. I also make sure that NO sourcecode is available in the library path, only .dcu/.res etc. Only browsepath has the sourcecode, and special circumstances are handled through searchpath for the project.
Just limit what you compile in any situation.
A few years later I am struggling again with increasing compiling times. I am currently using Delphi XE4 and I am at a point where I absolutely need to refactor the units references. I thought about a new way to identify where are the problems:
I’m using Process Monitor from Microsoft/SysInternals to monitor the compiler:
I start Process Monitor with a filter to show only dcc32.exe
(or bds.exe when working from the IDE).
I build my project from the command line.
At the end I look at the CreateFile operations in the log of Process Monitor.
For each unit there will an entry for the .PAS file (when the compiler starts working on this unit) and one for the .DCU file (when the compiler is complexly done with this unit). By working on the log with a text editor and/or with Excel I can extract this kind of information:
A kind of “tree”, where you recursively see in which order the units have been compiled.
For each unit the delay between “.PAS file opened“ and “.DCU file written”.
Then I try to interpret the results to find places where doing some refactoring would speed the compile time. It is not so easy, but I’m getting some encouraging results.
Our team had been using Delphi 6 for many years, then switched to Delphi 2006 years ago. With both versions we have the following problem: frequently the compiler complains about a unit which is supposedly used recursively. This unit is a 40k LOC unit which is at the core of a project with almost 1 million LOC (third party included).
The error message is incorrect: a full build on the project always works. Unfortunately, the error message does not tell us where the supposed circular reference is, just the name of that unit. Sometimes it even happens that valid error messages are listed 2-4 times until that circular reference problem is "found". Clearly the compiler is running in a circle here. Because of the size of that project it is hard to find the problem manually. Therefore I made a tool which proves that there really is no circular reference (the tool creates a directed dependency graph of the units and determines the coherence components in that graph - there are none except if I deliberately put some in).
This is not only affecting F9 compilation but also code completion / insight which is not working most of the time. Sometimes it works when I press ctrl-space a second time...
Any ideas how we can isolate or even fix the problem? Note that it will be very hard to split the 40k LOC unit into smaller ones because it contains about 15 large classes which depend on each other in the interface section (I know it's bad but should work anyway).
Update
We are constantly refactoring but this is one tough unit to refactor because everything depends on everything, almost. Have been trying to get around it via interfaces but we are talking about some classes with 100s of methods and properties. And it would be slower.
Upgrading to D2009 may be an option down the road but right now we're stuck with D2006 (the unicode stuff and the price tag are two of the stoppers here). Question is anyway if it would help since the problem is in there since D6 at least.
About trimming the uses clauses, we have been doing this frequently with Icarus. But that did not help so far. We are down to 90 custom units in the interface section now. However, with a true circular reference the problem could be in any unit. Also tried to add all units to the dpr.
The project shares a lot of code with other projects, and there are some IFDEFs. However, the defines are not set up in project options but via a common include file. Therefore all modules should see the same defines. Also, the problem reoccurs shortly after a full rebuild without switching to another project.
I will probably be downvoted for this. In D2005 I had a 10k loc unit (datamodule) that flat out stopped compiling. Had to separate out some datasets/code to another datamodule. That 10k unit was and is a mess. You really should consider refactoring out some code to other units. My module has since D2005 / separation grown even worse, but it still compiles in D2007. So my answer is a) refactor and b) upgrade to D2009.
It seems clear that this is due to a slight difference between the background compiler and the real thing. You could look around (QualityCentral) what's known on that topic.
Also, since you didn't explicitly state this, you should remove unnecessary units and move uses statements down to implementation if possible. Maybe your tool could help with this.
And just to be sure you should check the unit aliases and Path settings.
You write that a full build does always succeed, but shortly after the incremental build fails with this error. Assuming that you experience this in the IDE, have you tried to use the command line compiler dcc32 to do incremental builds?
If you don't feed it the "-Q" switch (which probably most Makefiles or scripts for command line builds do) it will output a lot of information what files it compiles in what order. You could either try to do an incremental build after the error appeared in the IDE, or you could keep a command line open next to the IDE and Alt+Tab to it for compilation, skipping compilation in the IDE completely.
I simply assume you have a way to build using dcc32, one way or another - with the size of your project I can't imagine otherwise.
We regularly fall in similar problems, and we never managed (or bothered long enough) to find the precise cause. There seems to be a problem in the order which Delphi chooses to compile the units when hitting Ctrl-F9, which is incompatible with the actual dependency order of the units.
Have you tried deleting "MyBigFatUnit.dcu" before hitting Ctrl-F9?
Have you tried to re-order the declaration of your units in your dpr/dpk files, so that units appear in a correct compilation order? (i.e.: if unit B depends on unit A, unit A should appear first in the dpr/dpk)
Do you have any other projects that use part of the same codebase? If you compile one of them under different compiler settings or IFDEFs, it might change certain things in some of the DCUs which would lead to a circular dependency. A full build rebuilds all DCUs and then the problem goes away.
Try Icarus (free) from Peganza. If that does not tell you what the problem is, try their Pascal Analyzer.
We have this problem as well, also with a fairly large codebase.
We are currently using D2009, but have had this problem with all previous versions of Delphi.
It most frequently happens immediately after doing an update from source control, so I suspect there is some timestamp issue within the Delphi build process.
In our case, if Ctrl-F9 fails and reports the circular reference, a second Ctrl-F9 will generally work
A way I have been told to deal with this is to open another arbitrary file in the project, change that file, save it, and then try running the incremental compile again. Surprisingly enough, this usually works.
We have a 4 MLOC project where this comes up from time to time and this "solution" works for me.
I've fought this before, in my experience the error is quasi-legitimate. It's been a quite a while (the error has nothing to do with the version) but my memory of the situation is that it involves a loop in which part of the loop is in the implementation.
Unit A uses B in the implementation. Unit B uses A in the interface. If you compile B first it calls for A but since the call for B is in the implementation it succeeds. If you compile A first it calls for B, B turns around and calls for A in the interface, boom. Such loops are only safe if both cross references are in the implementation.
The solution is to design things so there is a minimum of stuff used in the interface and to make certain there's nothing resembling a loop in those units. So long as you keep your type definitions separate from units with code this is pretty easy to do.
The error coming and going depending on what you are doing is a hallmark of this issue as it comes down to how you enter the loop. When you do a full build the order is consistent and you either get it 100% or 0%, it's not random.