I currently have code, in C#, that requires a BasicHttpBinding object to connect to SSRS. As it stands, I initialise this object using values assigned in code, rather than reading it from the app.config (this is because the platform I'm deploying to, MS CRM 2011, does not provide access to the app.config file for reading. In fact, I don't think the app.config file even gets copied to the server).
I'd like to make this binding editable without recompiling so the solution can be installed easily at different customers. The cleanest way I can think of is to have the binding config stored in a web-resource (for non-CRM people, this is just a name for a file stored inside CRM that you can access from code), but I'm not sure of the best way to parse that config into a BasicHttpBinding object? Manually parsing it and setting the properties seems inefficient and not very robust.
Is there any way of getting .NET to to it for me (similar to the BasicHttpBinding(string) constructor, but since I don't have access to app.config this isn't an option)?
Given the (presumably) low volatility of this configuration, could you not store it in the plug-in secure / unsecure configuration in your plug-in registration step? Accessing this at runtime would be significantly quicker that connecting to CRM to retrieve the contents of a web resource. Granted, it means that you require the plug-in registration tool and a sys admin to make changes but it also means that any sensitive data is obfuscated from users. App.config for a plug-in is indeed unavailable - but then that's what your config section in the plug-in step are for.
As far as using the constructor for BasicHttpBinding
It sounds like you're looking for a way have all of the constructor magic done for you by just passing the key name for a config section. I don't believe that this will be possible but in any case your suggestion that "Manually parsing it and setting the properties seems inefficient" is probably unfounded - after all that's exactly what the native constructor will ultimately be doing anyway. Hide it behind a function and you'll never know it's there ;)
I have connected to the SSRS web service in plugins previously (beware - not supported in CRM Online due to service not being exposed :( ) and I took my usual approach of storing config, as above, in the plugin configuration and then read it into an XmlDocument at runtime then parsed out the values as required to instantiate objects/set properties.
So long as the config is the same for all requests to the plugin, you can also potentially afford some efficiencies by setting the config values (not the connection itself) as class-level properties in your plugin (even though this is against advise in the SDK - but that's due to thread-safety which shouldn't be an issue with "static" config values such as this) and only read the values from config if they are not already set.
Related
This question already has answers here:
Pros and cons of AppSettings vs applicationSettings (.NET app.config / Web.config)
(6 answers)
Closed 7 years ago.
.NET has supported a number of ways to store configuration settings for quite some time, which has led to a good deal of confusion. It doesn't help that the top three hits for a search for "Settings in C#" over on MSDN are ten years old (and many of the related questions [2] on this venerable site are also 6-7 years old). It is really hard to determine which of the early recommendations have since been deprecated.
In my project I have chosen to use ApplicationSettings (as opposed to AppSettings) because:
That appears to be the current recommended approach on MSDN (and here).
The AppSettings element appears to be deprecated as of .NET 4+* (topic is "no longer available"; but see "Other Versions" on that page.) [*Although, confusingly, the AppSettings Property is still supported]
I prefer to edit the settings in a designer tool, rather than an xml editor.
I prefer strongly typed settings.
However, now I need to decide where to put my connection strings. The most current documentation on MSDN (and here on SO, as well as CP) still appears to recommend using the <connectionStrings> section of the app.config. But accessing this from the code requires the use of the old syntax, which I now consider obsolete along with appSettings. In other words, it makes no sense to read one part of the app.config file with the old syntax:
ConfigurationManager.ConnectionStrings["MydDBConnName"];
and another part of the same file with the new syntax:
Properties.Settings.Default.myOtherSetting;
Plus, it prevents me from being able to edit the strings in the designer.
So bottom line: is there any reason not to standardize all my configuration settings (including the connection strings) in the ApplicationSettings element?
The ConnectionStrings section allows you to not only define the connection string in the config, but also lets you choose the provider, so you code can (theoretically) use any subclass of DbConnection, DbCommand, etc.
In reality, however, supporting that kind of flexibility means you have to use SQL statements that are provider-agnostic (meaning you can't do things like date math that do not have a standard SQL syntax), and require more "plumbing" code to create the right types of objects. You can see some examples here.
If you only support one database provider (SQL Server, Oracle, ODBC, OleDB) then there's no real benefit to using ConnectionStrings over a string application setting.
I suggest that you keep your classes settings-source agnostic.
For example if you have a DatabaseContext that requires a connection string, then inject that primitive dependency in the constructor. Do not locate the connection string directly via ApplicationSettings.
Locating primitive dependencies (e.g. settings) from your classes is exactly the same as using the Service Locator anti-pattern.
The only location in your application that should acquire settings (e.g. connection strings) is the Composition Root. So you can obtain settings from ApplicationSettings here and inject them into your classes.
This allows you to change your mind later if you want to use a different way for storing/retrieving settings.
As you must have read on the pages that you linked to, the main benefit of using <connectionStrings> is that it provides mechanisms for encrypting the strings in order not to keep passwords in clear text. If you use Windows authentication to connect to the database then I guess you don't need it and really doesn't matter where you keep your connection strings. It's just a standard way of doing this.
I believe, however, that you are mistaken saying that the 'old syntax' is deprecated. For example, <appSettings> is still documented, it just changed the address. It would bring havoc if it was. It's not my area, but I think what you refer to as the 'new syntax' is the way of accessing settings in a desktop application, you don't have it in server-side applications.
I believe the ApplicationSettings element is just used for organization. If you notice in Entity Framework or most other database connections, they store them in the config under the ConnectionStrings element. The only thing I would worry about is storing any sort of sensitive data such as connection user and password. A common way to get around that is to allow the Windows Authentication to handle the connection.
A standard is what you make. For example in my current work environment the only information that changes in a particular application is the server name. We have dev/test/prod servers. Therefore we only store the SQL Server name in the configuration file. The database name doesn't change. We just read the database server from the configuration file and build the string in the application.
I'm building out my application and I'm at a point where I've hardcoded a lot of settings at the top of my class files - stuff like ApiSid and ApiKey, SmtpServiceUsername, MyEmailPassword etc. I'm now trying to consolidate these and I see two options:
1) Push them all into web.config. I don't like the thought of muddling up my web.config with tens (almost 100) settings though... I also feel uncomfortable with security here.
2) Build a static class that just contains these settings (Settings.cs) - basically housing a bunch of constants that are referenced throughout the app.
I feel more comfortable with the second approach because I can keep my settings totally isolated and not worry about exposing them via web.config - is there anything inherently wrong with this approach?
is there anything inherently wrong with this approach?
What makes you think putting constants in the code is any more secure than in the config? The compiled DLLs are right there next to the Web.Config, if somebody can examine one of them they can examine the other one. Hard-coded values can be de-compiled pretty easily.
Config files exist for a reason. Specifically, if any value is going to change per environment then it belongs in the config file. That way the same identical codebase can be used in any environment (development, test, production, etc.) and you'd just edit the config values for that environment. Having to re-compile the code just to deploy the same version to a new environment is less than ideal, since it's no longer the same version.
I don't like the thought of muddling up my web.config with tens (almost 100) settings though
Why not? If they're all flat static values, a list of appSettings keys would be fine. If there's more structure to them, create custom config sections.
This is not necessarily the best approach but I'd store these kind of settings in the database. This gives you database security for the settings plus it's easy to update the settings without having to stop / restart the application so you avoid kicking out users.
Once you have your settings in the database, you can load them periodically (like every 15-20 minutes) to detect changes. In the meantime, create a dictionary of the data and either wrap it in a class that provides type-safe access through properties or just use the dictionary directly. Since this is web application, you'll have to use a thread-safe class (like ConcurrentDictionary) to make sure multiple threads can safely access your settings.
If you have so many settings, web.config would be cluttered and every change would force an app pool restart. As #David mentions in his answer, the config file gives you an easy way to have different settings for different environments but this is also easy to do with a database approach where settings may be present once per environment.
In my IIS I want to be able to load some variables from an external file.
Reading them from the web.config is a possibility but if I will want to update the variable without restarting, it will require me to edit all the web.config files in the cloud.
Reading them from a centralized db is also an option but some of my apps dont require a connection.
What is a recommended way to manage application variables for IIS.
Thanks.
The recommended way is a Web.config file, that's why there is a built-in appSettings section. Now, as far as updating variables without restarting, you wouldn't have to edit them in the cloud at all. Have a copy of those Web.config files locally, in a testing environment that's like PROD, and then update them there first. Then you can test those changes and upload the entire Web.config file which will cause the application pool to reset automatically.
EDIT
With more information now available because of the OP's comments -another good solution may be to leverage the machine.config instead of the web.config for those settings that needs to be changed across multiple websites on the same server, especially if it's time sensitive that the applications see the change.
Since you don't want to have an application restart (which both web.config and machine.config will do), I would recommend keeping it to custom settings object that can be responsible for updating itself independently. Consider a serializable object with settings that appear in a NameValueCollection (similar to a web.config or app.config file). This would allow you to add settings similarly. Some steps to consider:
Add a shared directory to a common network location so that you only have 1 copy of the file itself for all websites. Something like \\somecommonserver\shared\configuration\.
Add a static application variable to the code base so that each application pool will have a record of the last date/time the file was updated --> Application["CustomConfigLastUpdate"] = configFileInfo.LastWriteTimeUtc; (I recommend using UTC because servers time may differ, but UTC does not)
Each time the settings are accessed may be a bit too heavy, but at some determined interval (each access, every 5/15/60 minutes, whatever) check if the configFileInfo.LastWriteTimeUtc property is greater than what is stored in the application variable, then you need to go get a fresh copy of the settings.
Create your custom object with the [Serializable] attribute and provide it a LoadFromXmlFile method that receives a filename as an argument whose responsibility is to repopulate itself.
I have done this before to achieve a similar goal, but I do not have those code samples here at my office. Here is a good SO question that has a relevant answer with code describing similar behavior: How to Deserialize XML document
I am looking at ways to make our application more extensible and easier to manipulate without having to alter the web.config (or, in our case, application.config files, which contain the appsettings node).
One way I have thought about is keeping the app settings in the database table that has a sqlcachedependancy. This means that:
Any time a setting is changed in the database, the cache is invalidated, and the settings are retrieved again, thus updating the application in realtime without having to alter files and restart the entire app.
We can create a custom tool which allows us to alter the settings.
The cons as I see it are that this may cause serious logic problems in that, if you have something that checks an appsetting at the start of a process, and it then changes halfway through, you could end up unintentionally altering the process flow, as the requirement for a complete application restart is negated.
Is there a way round this?
Is there a better way to manage appsettings, so that you can alter them on the fly remotely for one, several, or all servers in one go?
I think you've nailed the two major players:
either you have access to the file system and you put all your settings in a plethora of *.config files there
OR:
you don't have access (or only very limited access) to the server's file system and thus you're probably better off putting config settings and user preferences in a database, basically leaving nothing but the connection string to the config file on disk
Both approaches have their pros and cons. I've been trying for a long time to find a way to "materialize" a config section from a database field, so that I could basically just use the config XML, but stored in a database field. Unfortunately, the entire .NET 2.0 config system is very much "locked down" and just only assumes data will come from files - there's no way to plug in e.g. a database provider to allow the config system to read its contents from a database field :-( Really too bad!
The only other approach I've seen is a "ConfigurationService" in the StockTrader 2.0 sample app provided by Microsoft, but for my needs, it felt like overkill and like a really complex, really heavy-weight subsystem.
You could use SQLite, which will be a self-contained DB in a single file. Two birds with one stone?
If you reference an external config file that contains appsettings (leaving everything else in the normal app.config) then I believe editing it only reloads those settings, it doesn't force the whole app to restart.
There's a similar question on the subject here:
Nested app.config (web.config) files
WRT the problem of values changing in the middle of program execution, I guess you could locally cache the values, and raise an event when they change, allowing routines to reach a suitable point before using the updated values.
I think in asp.net we sort of get this for free because each page lifecyle is distinct, so the value is simply applied to new page requests only, not in the middle of an execution.
Edit: A little extra info:
Configuration Changes Cause a Restart of the Application Domain
From MSDN:
Changes to configuration settings in Web.config files indirectly cause the application domain to restart. This behavior occurs by design. You can optionally use the configSource attribute to reference external configuration files that do not cause a restart when a change is made. For more information, see configSource in General Attributes Inherited by Section Elements.
More information on the ConfigurationManager class in the System.Configuration namespace which could be used to modify the config files programatically (ie in a custom tool, if relevant disk read permissions can be provided). If you stick to using the built in configuration classes, I think changing the external configs, would not cause application restart, but would raise events (such as property changed) which you could handle, to ensure your code is not caught out by changing settings.
Is there a way at runtime to switch out an applications app.config (current.config to new.config, file for file). I have a backup/restore process which needs to replace its own application.exe.config file. I have seen this post but it does not answer how to do this at runtime.
Turns out I can swap the .config file for the new one and do a ConfigurationManager.RefreshSection(...) for each section. It will update from the new .config file.
Microsoft .NET's app.config is not designed for your scenario, as well as many others. I often encounter a similar need, so I have spent a lot of effort designing a solution.
Redesign to use app.config only as a configuration bootstrap: specify where to find the rest of the real configuration data. This information should almost never change, so there is no need to handle file watching or application restarts.
Pick an alternate location for the real configuration data: a file, a database, perhaps even a web service. I prefer a database most of the time, so I create a configuration table with a simple structure that allows me to store my data.
Implement a simple library to wrap your configuration access so that you have a simple API for the rest of your application (via dependency injection). Hide the usage of app.config as well as your real configuration storage location(s). Since .NET is strongly-typed, make the configuration settings so--convert each string retrieved into the most-specific type available (URL, Int32, FileInfo, etc.).
Determine which configuration settings can be safely changed at runtime versus those that can't. Typically, some settings need to change along with others, or it simply makes no sense to allow them to change at all. If all your configuration data can safely change at runtime, then that makes things easy, but I HIGHLY doubt such a scenario. Hide the changeability and interdependencies of the configuration settings to the extent possible.
Design the response to the unavailability of your real configuration data. I prefer to treat the absence of any configuration setting as a fatal error that aborts the application, unless I can identify a usable default. Likewise, I abort in the absence of the configuration storage container (file, database table, etc.).
Enjoy, and best wishes.
Are you able to restart the application when you detect that you need to switch files? If so, it's just a matter of switching the files and restarting. Now, the tricky bit is if .NET keeps the app.config file open while the program is running. I suspect it doesn't, but if the most obviously approach fails, I suggest you have a second application (cfgswitcher.exe) which waits for the process with a PID specified on the command line to terminate, then switches config files and relaunches the original process. Then your app would just need to launch cfgswitcher.exe (passing in its own PID as a command line argument) and terminate.
As I say though, it's worth trying the more obvious approach first.
EDIT: If you can't restart the application (or even part of it in a new AppDomain) then various aspects of app.config (assembly bindings etc) can't be changed. If you're only interested in your own configuration sections changing, then I suggest you store them in a separate config file and reload them whenever you want to.
Look at the events available to you on the ApplicationSettingsBase class. There are PropertyChanged & SettingChanging that may give you what you need.
You could also watch the file and if it has changed call the reload method to get the new settings.
I don't think it is possible at all to switch the configuration at runtime without restarting, so if you can't apply Jon's approach, you should try to come up with an other approach.
Anyway, maybe it's just me not having enough information about your scenario, but this kind of feels fishy.
Are you sure that swapping the configuration file is the best way to achieve whatever requirement you need to meet? I mean, this is quite an uncommon thing. If I were you, I would try to come up with some other approach.