We have a lot of products that are saving their "states" on the registry.
What is the best practice on saving program states? What are the advantages/disadvantages of saving program states as a registry entry or saving program states to a flat file such as XML?
Thanks!
The obvious awswer would be that storing those states in a normal file, makes it easier for users to backup/restore the state manually.
Also consider that the registry has some keys that are special for each user in the system.
I think registry is the best option to store user-specific information that can be discarded and recovered easily (eg, the last username used to login). Other data should be in a settings file that can be backed-up.
For years programmers had their app settings stored in config files. Then the times changed, and for years they used the registry instead - many of them used it badly, and it caused issues when Vista and its UAC came on the scene.
Nowadays, especially in the .Net world, Windows developers are moving back to storing stuff in config files again. Personally i think that is the best way, if you need to move your app to another machine, or reinstall your OS, all you have to make sure you do is save your config file to retain your settings.
There are things that you may still want to store in the registry though, such as (encrypted) licencing info. For everything else, config files are good. Do pay attention to UAC and file virtualisation though, so that you don't run in to trouble further down the track.
Personally I'd go for the flat file.
(I am assuming that "registry" means windows registry?)
A flat file allows you (or even the user) to inspect and eventually even modify manually the values.
Depending on your situation this could be helpful for debugging, repairing mis-saved data etc.
Unless you thing you want to have the data to be "opaque" and therefore "hard to find/manipulate", the registry offers little in terms of benefits. Maybe it's faster, but if you have lots of state to save you better use an embedded DB instead of a flat file.
I used to follow Redmond doctrines. My programs used .INI files. Then I dutifully switched to the registry - and users started complaining. So, I bucked the trend and switched back to .INI files.
Some want to edit them (good/bad?). Some want to back them up, or transfer to a new machine. Some don't want to lose them if they reinstall windows.
AS a user, I have multiple partitions. Windows/programs/data/swap (and a few others). No programs go onto c:\program files, they all go into the programs partition. No data which I can control goes into c:\user data, it all goes into the data partition (use tweakui power toy, or regedit to change the defaults (but not all programs are well behaved and read the registry for those paths - some just hard code them)).
Bottom line - when Windows gets its panties in a fankle, I do a total re-insatll (approx every three months), and I format the C: drive.
By formatting the windows partition, I get a clean install. My data and programs are safe, though I may need to reinstall a few programs, which is why I go with portable versions where at all possible.
Imo, the registry is the biggest evil ever perpetrated on Windows - a single point of failure.
My advice? Locally stored config files. INI if the user is allowed to edit, serialized or binary format if not.
Or, you could offer a choice ...
Personally I go for a flat file, whether it's an INI file or XML file makes no difference to me. However in my line of work, we've had customers prefer the registry instead due to issues relating to deployment. It depends on who your client base is, and what the person keeping your product working prefers.
I always use regular files because its much easier to develop =)
Simple io vs I don't remember how read/write registry
Simple file copy/paste vs export/import keys for backup/developpement multiple versions of config for testing
Note that all of these advantages also translate into deployment strategies and generic client usage of the configurations
Depends how heavy deployment is. Most of my applications are XCopy-Deployable, that is they don't need an installer and can just be copied/unzipped. So I use .ini Files (using my own INI File Parser as .net has no built in one)
However, if your application needs to be centrally manageable (for example, using Windows Group Policies) or if you have a "heavy" installer anyway, the registry is the prime choice. This is because Applications that are installed normally to to C:\Program Files, and normal users do not have write access to this directory. Sure, there are Alternatives (%APPDATA% or Isolated Storage which has to be used when the Application is a Silverlight app), but you can as well "go with the flow".
Of course, if your application is supposed to run on Mono, you can rule out the Registry anyway and should go Flat Files.
Related
I am a WordPress Designer/Developer, who is getting more and more heavily involved with using version control, notably Git, though I do use SVN for some projects. I am currently using Beanstalk for my remote repo.
Adding all of the WordPress files to my repo is no problem, if I wanted to I know I could .gitignore the wp-config file, but since I'm the only developer, currently, and these projects are closed source, it really makes little sense.
WordPress relies heavily on the database, as any CMS does, to keep textual content, and many settings depending on the specific plugin/theme configuration I'm using. I'm wondering what the best way of using version control on the database would be, if it's even possible. I guess I could do a SQL dump, though my MySQL server is running on Windows (read as: I don't know how to do it), and then add the SQL dump to my repository. But when I push something live, that poses huge security threats.
Is there an accepted practice of doing this?
You can backup your database within a git repository. Of course, if you place the data into git in a binary form, you will lose all of git's ability to efficiently store the data using diffs (changes). So the number one best practice is this: store the data in a text serialised format.
mysqldump is a suitable program to help you do this. It isn't perfect though. If anything disturbs the serialisation order of items (eg. as a result of creating new tables, etc.) then artificial breaks will enter into the diff. That will decrease the efficiency of storage. You could write a custom serialiser to serialise changes only -- but then you are doing the hard work that git is already good at. Just use the sql dump.
That being said, what you are wanting to do isn't what devs normally mean when they talk about putting the database in git. For instance, if you read the link posted by #eggyal (link to codinghorror) you will see that what is actually placed in git are the scripts needed to generate the initial database. There may be additional scripts, like those to populate the database data with a clean state, or to populate it with testing data. All such sql scripts are text files, and pretty much the same format as the sql dump you would get from mysqldump. So there's no reason you can't do it that way with your day-to-day data as well.
There are not many software available to version control databases like MySQL and MongoDB.
But one is under development and the beta version is about to be launched soon. Check out Klonio - Version Control for databases
The article How to Sync A Local & Remote WordPress Blog Using Version Control gives advice on how to automate sync between two instances (development, production) of a WordPress blog using Mercurial. Is mentions that for this scenario, Git and Mercurial are very similar.
Step 4 (Synchronizing The Databases) is of interest here.
The database content will be exported to a file that is tracked by the revision control. Each time we pull changes, the database content will be replaced by this file, making our database up-to-date.
Then, it elaborates on conflicts and the scripting part of the job.
There is a version control tutorial in Mercurial out there, if you're not familiar with it.
If you are only interested in schema changes under version control, there is a nice stuff SqlRog. It extracts schema into the project files that can be put under the git.
Be aware that Wordpress stores all news feed content in the database, so even if you don't make any changes, there will be a lot of changing content.
hope this is a good spot for my question,
for it i SW Related, but not code related.
We, in our company are using TRAC for Issue tracking and management of the Code links,
I am very satisfied by it, and like how it is working.
i have about several environments (1 per project) and every time we change a setting in the Configurations (e.g. Users & Permissions, Severity, Ticket types, etc...) we need to change all of them.
I Use
[inherit]
file=../../../sharedTrac.ini
and delete the shared parts from the file.
for the preferences, but i didn't find a way to share the Configurations.
this is bad for several reason and the head reason is that is "Bugs me !!!" :p
Can TRAC read its configurations from a central definition, and the data from a local DB?
EDIT:
I noticed all these configurations are in the .db file (sqlite file)...
Is there a Ready made tool to copy the configurations from DB to DB ?
or should i go ahead and analyse what should be copied and how ?
You're almost there. Note though, that local settings will always over-rule inherited ones, so you must delete them in your <env>/conf/trac.ini files to make central configuration effective.
Specifically to the part of configuration inside Trac db: No, there is no sync tool yet. Given that there was one for user accounts that is still a beta after years, there's not much interest. You should use the trac-admin command-line tool (as already advised here) or start to directly sync parts the db by means of own (Python) scripts or custom db syncronisation. For a start have a look at the Trac db schema.
You can try to do this through command line. Just call appropriate "trac-admin" command for each instance. Example one-liner to add user profile:
for D in */; do trac-admin $D session add username "Full Name" user#email.com ; done
What are the benefits of choosing either:
a file in the user's home directory, e.g. ~/.myapp/config
using the user's gconf database
I already know a couple of things; storing data in the user's gconf database makes it a little harder for the user to copy those settings out and onto another account or computer. On the other hand, is it true that using gconf would be easier as the underlying data storage facilities are already sorted out, without you having to write your own configuration file handler?
I'd go with GConf.
However...
Perhaps if you are worried about users copying their settings in and out of the program, you should create import and export settings functions into your program? A user could then potentially save those settings in cloud storage like Ubuntu One or Dropbox and simply import their desired settings into GConf using the utility that you provide, regardless of the machine (even if they prefer to run from liveCD). Either way, very clearly indicate to the user where the settings are being stored. Many (but not all) Ubuntu users are comfortable backing up and restoring those configuration files by themselves using rsync or another backup utility.
Go with GConf. It was designed to store configuration for programs, and that is what you should use it for. Why should you re-invent the wheel from scratch when you have a perfectly good wheel making device in front of you?
When I say "configure" I mean where to save those values that could change very often (constants values like taxes rates or something similar) and then when you need to change them you don't want to re-compile your application.
Where to save those values? Database? XML File? Flat File?
It depends on how often these change and who or what changes them. For some application specific settings, it's best to use an XML or config file, where the developers are the ones responsible for updating it. For other "businessy" values (like exchange rates, tax rates, etc), it's best to keep them in the database and provide a UI for users (not developers) to update.
It also depends on how many apps depend on this value, for example, if several applications depend on some setting (such as email server addres), it's best to put it in a database since it'll be easily accessible from any machine where the app is running.
I use INI files for potentially user-configurable files, and BIN files for data that save session state between runs.
But, it is very dependent upon what type of application you are developing.
it depends on how your app is architecture. you could design your app in such way that you could change the location of you configuration. by just injecting the provider.
Normally I use Ini files or XML if the data is structured.
For applications that already use a database and you don't want to have the user to change the data easily, you can use the database.
I almost never use binary data unless you want to obfuscate the data for the user.
Regardless of app, you're probably going to have at least 3 sources of configuration data:
Command line flags, usually for bootstrapping your run-time environment, e.g, finding config files, setting debug flags, include paths, class paths, etc
Config files, potentially more than one that may override each other. These usually boot strap your application: connection strings, cache settings, build-specific settings, etc
Control data in a database. Things like timezones, conversion rates, stable display values, etc. This data should also be versioned in the database (as in, a "Data Version" field, not living in a version control system). Versioning it will save a lot of headaches when you find you need to change a setting for a new release, but the old release will break if you change it.
Generally, anything that changes at run-time should go in the database. Anything that is sensitive and rarely changing should go into the config files, and any hacks should go on the command line (--[no]enable-bug-287438-hack can be very handy when you need it).
I prefer the simplicity of a flat ini file. Here's an example Setting class that you might find useful.
Would you recommend working with multiple programmers on an MS Access application?
One of our MS Access application has grown to the point where the number of changes (bug fixes) and new features can no longer be handled by one programmer in the requested time frame.
We are trying to introduce version control using the undocumented SaveAsText and LoadFromText procedures in VBA to make collaboration on this application possible. Unfortunately we have already run into problems loading modified forms and reports back into Access as a checksum is stored in every form text file.
Before putting time into building an import/export application to compile text files into an Access database, we would like to hear your recommendations.
I think you should avoid this path at all cost, and try and persuade management into redevelopment.
It's a bitter pill to swallow, but this is going to need to be redeveloped sooner or later, and you are just saving them time and money.
We were using Microsoft's own version control add-in for MS Access 2000/2002/2003 for about 5 years now, and I can't remember a single serious problem. Usability of this add-in barely deserves a "B", but it must be much, much more convenient than fiddling with any ad-hoc method involving manual or semi-manual exporting/importing of Access forms, modules, etc.
We were using VSS as a version control system all the time. No problems whatsoever. However, if you have some good reasons to avoid VSS, you may have some options:
The version control add-in that we were using does not require VSS. Theoretically it can be used with any version control system that implements Microsoft Source Code Control Interface (MSCCI). For example, when we had to let somebody work on this project remotely, we used SourceOffsite by SourceGear. Access version control add-in worked with this third-party product fairly well (not without some quirks, but well enough). So, if your favorite version control system complies with MSCCI, you could try to use it.
Now that Microsoft has this Team Foundation thingy, apparently there are other options to be used to integrate MS Access with version control. We did not explore this path, though. This article may be a good start for exploring it.
Hope this would be of some help. :-)
P.S. I am not a big fan of MS Access. In fact, I rather hate it as a platform for a user front-end. If I had a choice, I would run away from it yesterday. :-) However, I must admit that existence of this version control add-in is one of the few things that makes maintenance of our old Access+SQLServer project more or less tolerable. :-))
In addition to what I already said here, I should add that the whole system works very well. The comparison process takes less than 30 minutes a week, for a team of 3 programmers. So let's describe it a little bit.
We have basically 2 versions of our Access program:
The "Developer's version", with all the stuff in it.
We each begin to work with an identical version of our developer's edition. As each one modifies or add parts of the code, we have to run some comparison routine on a regular basis. To do so, we have an object-export routine to a common "comparison" folder. An object (module for example) is exported as a text file (saveAsText command, do not work with tables, see infra), it will be compared to the existing equivalent text files in the folder. If files are identical, there is no file exported. If files are different, the new module is exported with the developer's name as an addition to the file name (if modQueries.txt exists, then modQueries_philippe.txt is created...). Of course if there is no equivalent .txt file in the folder, it will be created at first export.
At the end of the period, we would get in our folder the following files
modQueries.txt, being the first "original", last common version of the module
modQueries_Philippe.txt, with Philippe's modifications
modQueries_Denise.txt, with Denise's modifications
As the module was not modified by other developers, their export did not lead to the creation of a specific modQueries_developersName.txt file
If for any reasons Denise exported many times her module, only the last version is in the comparison folder.
We can then compare (with a "text file" comparer) the different versions and create the "updated" version of the module. We have a screen giving us the number of objects in the comparison folder, number of version for each object, and it is even possible to open the file comparer directly from the developer's interface (We use "File Compare Tool" which has a command-line mode and can then be started directly from Access).
The forms compare issue is quite special, as one of our rules is to have no specific code in our forms (please see here for more details). Forms are then only for display, so usually we do not even compare them. We just make sure that each one of them is updated by only one person (which is quite logical).
The table compare issue (we have local tables) can be only made between mdb files. As we export one text file per module, we also export one mdb file per table. We have a small routine allowing us to identify table differences at the structure level or at the record level.
After each comparison procedure, a subroutine will use all the objects available ini the comparison folder and create a whole new clean mdb file from scratch. This is the new developer's version. Every developer can then copy it on his computer and continue his work.
Developer's versions do not have numbers, but contains last client version number.
The client version, with limited stuff, automatically distributed to users
Each developer has the possibility to build a "client" mdb for final users. This mdb is created from scratch, in a way quite similar to our developer's version, but not all objects are exported. Some specific switches are turned off (special keys, access to code, etc). This mdb holds a version number as a property. The version number is used to build the name of the mdb file.
At production time, this mdb file is zipped and placed in a specific "distribution" folder. Each time a user starts the app, it will automatically check this folder to see if a new version is available. If yes, the client mdb file is updated from the distribution folder, and the app is restarted.
This distribution folder is replicated at night time with our overseas agencies. Users abroad will then be able to install the new version on the following day.
Following the direction provided by Yarik we settled on continuing developing in Access using the Access Add-in Source Code Control, the SVN SCC API Plugin by PushOk Software and Subversion. This stack provides us with seamless Access integration, full-backup and restore and an open version control system.
We had to install a hotfix to Access 2003 and make sure the default database file type matched our database file type to make it work.
We will continue to update this answer with our findings.
Have look at this thread:
How do you use version control with Access development?
Sounds like a terribly painful way to do team development. If you have any options for porting to another environment like VS2008 that would be my recommendation.
There is no easy way to work on Access as a team and even version control might be a bit tricky.