I am developing a transactional application in .NET and would like to get some input on how to properly encapsulate database access so that:
I don't have connection strings all
over the place
Multiple calls to the same stored
procedure from different functions
or WORSE, multiple stored
procedures that are different by a
single column
I am interested in knowing if using an ORM like NHibernate is useful, as it may just add another layer of complexity to a rapidly changing data model, and artifacts need to be produced on a tight schedule.
I am more interested in methods or patterns OTHER than ORM packages.
There are at least two widely accepted design patterns used to encapsulate data access:
repository (DDD)
DAO (Data Access Object)
For the sake of completeness I suggest you these books:
Patterns of Enterprise Application Architecture (Fowler)
Domain Driven Design (Evans)
If, as it appears, this is an important project and the DAL is a major risk factor, get someone involved who has done it before. You're exactly right that there are too many ways to run off the rails by trying to get this right the first time without solid experience.
There are any number of patterns for accomplishing this, but I'd look for someone who has a simple set of well-defined patterns they are fully comfortable with.
as stated above, check out the repository and unit of work patterns. the books by fowler and evans are highly recommended. so is karl seguin's reader which gave me a cooler introduction to the just mentioned books. grab it at http://codebetter.com/blogs/karlseguin/archive/2008/06/24/foundations-of-programming-ebook.aspx
As Java Developer, i could suggest to read about jdbc templates, despite of it s'n not .NET you could learn how Spring framework encapsulates data access tier and get some ideas.
Related
I am developing an application which at the moment queries a (rather large) database via ADO.NET and hard-coded SQL statements. Admittedly this is ugly (i.e. no compile time errors thrown if a mistake is made in the SQL) and potentially dangerous (due to SQL injections, etc although this is unlikely to be a problem for this particular application) but this wasn't considered initially because this application is really only interested in a very small subset of tables in this database (at least for now...).
LinqToSQL seemed interesting but because this application is required to have the ability to connect to Oracle databases as well, that plan was a non-starter.
Is a project like mine suitable for integration with an ORM framework or would that be overkill?
I think an ORM should always at least be considered.
But it doesn't sound like you're even using business objects (Sometimes referred to as a Data Access Layer or DAL) which greatly undermines the usefulness of an object oriented language. I would address this first. If you find it's too time consuming to create all the CRUD for the business objects it's time for an ORM...
My personal favorite is nHibernate. Big learning curve but definitely worth it.
I would recommend a generated DAL instead of an ORM, or Linq.
Look into subsonic http://subsonicproject.com/. It is an open source DAL generator that is very easy to learn and use, and has a very low overhead.
I would definitely say that it is a candidate for an ORM framework. The overhead of setting up the ORM is quite small once you have familiarized yourself with a framework, and the benefits are many.
As you say, LinqToSQL is not appropriate if you might need Oracle support, but most other frameworks support Oracle.
If you only use a small subset of the tables, then you will only have to map a small subset of the tables and hence the setup cost will decrease even further.
Good luck!
Try using something that generates sql (Like Linq, only with Oracle), instead of an orm.
Why? Jeff Atwood explains.
Quote:
"At first you're like "whee! objects!" and then you realize-- hey, this is a lot of tedious, error-prone mapping code I didn't have to write before... "
I am in the processes of replacing the framework for a fairly complex business web application. Our application runs on a LAMP platform and the new framework will be an extension of CodeIgniter. In my research for framework design I decided to look into ORM, I have never done ORM before and I wanted to know if it would be valuable for our application. Then I stumbled on a very interesting blog post entitled "Why I Do Not Use ORM." This blog seemed to confirm many of my worries about using ORM and it also presented a solution similar to what I was already planning.
By "data dictionary" I plan to use this definition from "The Database Programmer" blog:
The term "data dictionary" is used by many, including myself, to denote a separate set of tables that describes the application tables. The Data Dictionary contains such information as column names, types, and sizes, but also descriptive information such as titles, captions, primary keys, foreign keys, and hints to the user interface about how to display the field.
So in choosing a "data dictionary" over ORM I may be exhibiting confirmation bias, regardless here are my reasons for being weary of ORM:
I have never used ORM before, I don't know much about it.
This framework needs to be built rather quickly, my boss has little time and I need to produce a working application that will allow for a smooth upgrade to a more modern framework.
My boss already thinks that I am over engineering this framework (trust me, I am no where close to that) and is paranoid about the framework preventing us from being able to do things that we need to, and creating bugs that we can't solve in the required amount of time. So far I have done a poor job of convincing him that change is good, I am not a very effective salesman and while the other developers can help me the boss still needs a lot of assurance.
Our old framework is procedural, our code is PHP, and our developers know SQL very well. ORM would be a big change.
Our database has dozens of tables, many with hundreds of thousands of entries running on a fairly old server. In the past we have been burned by code that repeated polls the database in a loop instead of doing one query to pull all of the needed data at once. Avoiding this problem with hand coded SQL is rather straight forward. Ensuring that this always happens where necessary with ORM is a huge unknown to me and appears to be risky.
Regardless, the solution of the data dictionary seems very promising to me as this blog post "Using A Data Dictionary" seems to provide a lot of useful features and some that are requirements of the new framework. Here are my reasons for preferring the data dictionary solution:
Implementing access control rules on the table rows themselves would be invaluable.
Auto-generating database changes, documentation, and schema checking would also be useful.
One requirement of the framework is a generic data history/auditing feature that can be applied to any sub-feature within our application. A data dictionary or an equivalent is essentially required to provide such a feature. The history must have detailed information about the structure and data types within the database.
Our systems hold a wide variety of data types that would more properly addressed if they treated as formal types within the application. For one, HTML fragments (of which we have many in our data, they are required) need to be encoded as entities in some cases, decoded as HTML in others, parsed for links and images in some cases, and always validated for correctness. Then there are dates, measurements, currency, and various other fields that could benefit from having a clear definition in the code of how this data should be manipulated.
The data dictionary idea that I would like to implement would be series of objects in separate PHP files, and there will be plenty of OOP, however it will be used as in a manner very similar to the data dictionary concept presented in "The Database Programmer" blog. It would be the single source definition of the complete database schema for the entire framework.
Now my question is, am I overlooking the value of ORM or is this a case where a data dictionary is the right tool for the job?
I think your question would be more interesting if you were making an initial architectural decision rather than refactoring an existing application. I don't see a single assertion in your question that suggests a problem that designing in an ORM would address; but several it would create. If two major stakeholder groups (owner and other developers are more comfortable with a more conventional design, it seems to me that an ORM would be swimming upstream.
I can imagine the (possibly undeserved) approbation that would be associated with the ORM as soon as a query is slow or transaction locking problems start emerging. Not to mention the impact on the development schedule. Why create an unquantified risk factor?
Do you have a framework which supports building applications using a "Data Dictionary"? If so, give it a try, it might solve your problems. If you haven't, then there are lots of good and working ORM frameworks out there which have large communities, which come with source (so you can fix bugs yourself even if the "vendor" refuses to help you).
If you want to get a quick glance at a nice web based ORM framework, I suggest Django or TurboGears. They are based on Python which will be a nice change after using PHP. I usually prefer TurboGears but Django seems to be more smooth at the moment. Both are easy to set up and you should be able to build a prototype in a day or two. That will give you an idea of the odds.
PS: I also don't think ORM tools are TEH SOLUT10N. I use Hibernate or SQL Alchemy when it makes sense but I often roll my own simple mappers.
I think that you have made a very good analysis for you situation. You know why you choose the Data Dictionary approach. So go for it.
Later on you might reconsider. If so, then there should be not a problem to use the Data Dictionary and a ORM for new developments in parallel. Both technologies are not mutual exclusive.
If you don't like the idea of mixing different technologies: Stick to a solid OOP design and separate concerns between domain logic and data access cleanly, then switching to an ORM shouldn't be that painful or at least possible.
Good luck!
I am continuing to delve into Erlang. I am thinking of starting my next web project using Erlang, and at this stage the only thing I will really miss from Ruby on Rails is ActiveRecord.
Is there a good alternative technology for Erlang?
Update:
The closest I have come to a solution is to ErlyDB, a component of ErlyWeb.
ErlyDB is a database abstraction layer
generator for Erlang. ErlyDB combines
database metadata and user-provided
metadata to generate functions that
let you perform common data access
operations in an intuitive manner. It
also provides a single API for working
with different database engines
(although currently, only MySQL is
supported), letting you write portable
data access code.
Well, the major advantages of ActiveRecord (as I see it) are:
You can persist your objects in a relational database nearly transparently.
You can search the database by any attribute of your objects.
You can validate objects when persisting them.
You can have callbacks on deleting, updating, or inserting objects.
With Mnesia:
You can persist any Erlang data absolutely transparently.
Using pattern matching, you can search the database by any attribute of your data or their combination.
QLC gives you a nice query interface for cases when pattern matching isn't enough.
No solutions for validating and callbacks, however...
So, what else do you have in ActiveRecord that is lacking in Mnesia?
I don't think there really is at the time of this writing. That may be because the kinds of systems being written in erlang and the type of people writing them don't really call for Relational Databases. I see much more code using mnesia, CouchDB, Tokyo Cabinet and other such alternative database technologies.
That's not to say someone might not want to create something like active record. It's just hasn't really been a need yet. Maybe you will be the first? :-)
You might be interested in Chicago Boss's "BossRecords":
http://www.chicagoboss.org/api-record.html
They are quite explicitly modeled on the Active Record pattern, and use a lot of compiler magic to make the syntax squeaky clean. BossRecords support save/validate as well as has_many/belongs_to associations. Attributes in your data model are made available through generated functions (e.g. "Employee:first_name()").
Some googling reveals libs / clients / wrappers for Couchdb described "ActiveRecord like libraries like CouchFoo", and advise to steer clear:
http://upstream-berlin.com/2009/03/31/the-case-of-activerecord-vs-couchdb/
http://debasishg.blogspot.com/2009/04/framework-inertia-couchdb-and-case-of.html#
as to your comment on "not suited for web apps yet", I think the pieces are there: mochiweb, couch, yaws, nitrogen, erlyweb. There's some powerful tools, very different paradigm, certainly, from rails, django and PHP.
I come from a java background.
But I would like a cross-platform perspective on what is considered best practice for persisting objects.
The way I see it, there are 3 camps:
ORM camp
direct query camp e.g. JDBC/DAO, iBatis
LINQ camp
Do people still handcode queries (bypassing ORM) ? Why, considering the options available via JPA, Django, Rails.
There is no one best practice for persistence (although the number of people screaming that ORM is best practice might lead you to believe otherwise). The only best practice is to use the method that is most appropriate for your team and your project.
We use ADO.NET and stored procedures for data access (though we do have some helpers that make it very fast to write such as SP class wrapper generators, an IDataRecord to object translator, and some higher order procedures encapsulating common patterns and error handling).
There are a bunch of reasons for this which I won't go into here, but suffice to say that they are decisions that work for our team and that our team agrees with. Which, at the end of the day, is what matters.
I am currently reading up on persisting objects in .net. As such I cannot offer a best practice, but maybe my insights can bring you some benefit. Up until a few months ago I have always used handcoded queries, a bad habit from my ASP.classic days.
Linq2SQL - Very lightweight and easy to get up to speed. I love the strongly typed querying possibilities and the fact that the SQL is not executed at once. Instead it is executed when your query is ready (all the filters applied) thus you can split the data access from the filtering of the data. Also Linq2SQL lets me use domain objects that are separate from the data objects which are dynamically generated. I have not tried Linq2SQL on a larger project but so far it seems promising. Oh it only supports MS SQL which is a shame.
Entity Framework - I played around with it a little bit and did not like it. It seems to want to do everything for me and it does not work well with stored procedures. EF supports Linq2Entities which again allows strongly typed queries. I think it is limited to MS SQL but I could be wrong.
SubSonic 3.0 (Alpha) - This is a newer version of SubSonic which supports Linq. The cool thing about SubSonic is that it is based on template files (T4 templates, written in C#) which you can easily modify. Thus if you want the auto-generated code to look different you just change it :). I have only tried a preview so far but will look at the Alpha today. Take a look here SubSonic 3 Alpha. Supports MS SQL but will support Oracle, MySql etc. soon.
So far my conclusion is to use Linq2SQL until SubSonic is ready and then switch to that since SubSonics templates allows much more customization.
There is at least another one: System Prevalence.
As far as I can tell, what is optimal for you depends a lot on your circumstances. I could see how for very simple systems, using direct queries still could be a good idea. Also, I have seen Hibernate fail to work well with complex, legacy database schemata, so using an ORM might not always be a valid option. System Prevalence is supposed to unbeatingly fast, if you have enough memory to fit all your objects into RAM. Don't know about LINQ, but I suppose it has its uses, too.
So, as so often, the answer is: know a variety of tools for the job, so that you are able to use the one that's most appropriate for your specific situation.
The best practice depends on your situation.
If you need database objects in table structures with some sort of meaningful structure (so one column per field, one row per entity and so on) you need some sort of translation layer inbetween objects and the database. These fall into two camps:
If there's no logic in the database (just storage) and tables map to objects well, then an ORM solution can provide a quick and reliable persistence system. Java systems like Toplink and Hibernate are mature technologies for this.
If there is database logic involved in persistence, or your database schema has drifted from your object model significantly, stored procedures wrapped by Data Access Objects (with further patterns as you like) is a little more involved than ORM but more flexible.
If you don't need structured storage (and you need to be really sure that you don't, as introducing it to existing data is not fun), you can store serialized object graphs directly in the database, bypassing a lot of complexity.
I prefer to write my own SQL, but I apply all my refactoring techniques and other "good stuff" when I do so.
I have written data access layers, ORM code generators, persistence layers, UnitOfWork transaction management, and LOTS of SQL. I've done that in systems of all shapes and sizes, including extremely high-performance data feeds (forty thousand files totaling forty million transactions per day, each loaded within two minutes of real-time).
The most important criteria is destiny, as in control thereof. Don't ever let your ORM tool be an obstacle to getting your work done, or an excuse for not doing it right. Ultimately, all good SQL is hand-written and hand-tuned, but some decent tools can help you get a good first draft quickly.
I treat this issue the same way that I do my UI design. I write all my UIs directly in code, but I might use a visual designer to prototype some essential elements that I have in mind, then I tear apart the code it generates in order to kickstart my own.
So, use an ORM tool in any of its manifestations as a way to get a decent example--look at how it solves many of the issues that arise (key generation, associations, navigation, etc.). Tear apart its output, make it your own, then reuse the heck out of it.
Is LINQ a kind of Object-Relational Mapper?
LINQ in itself is a set of language extensions to aid querying, readability and reduce code. LINQ to SQL is a kind of OR Mapper, but it isn't particularly powerful. The Entity Framework is often referred to as an OR Mapper, but it does quite a lot more.
There are several other LINQ to X implementations around, including LINQ to NHibernate and LINQ to LLBLGenPro that offer OR Mapping and supporting frameworks in a broadly similar fashion to the Entity Framework.
If you are just learning LINQ though, I'd recommend you stick to LINQ to Objects to get a feel for it, rather than diving into one of the more complicated flavours :-)
LINQ is not an ORM at all. LINQ is a way of querying "stuff", and can be more or less seen as a SQL-like language extension for different things (IEnumerables).
There are various types of "stuff" that can be queried, among them SQL Server databases. This is called LINQ-to-SQL. The way it works is that it generates (implicit) classes based on the structure of the DB and your query. In this sense it works much more like a code generator.
LINQ-to-SQL is not an ORM because it doesn't try at all to solve the object-relational impedance mismatch. In an ORM you design the classes and then either map them manually to tables or let the ORM generate the database. If you then change the database for whatever reason (typically refactoring, renormalization, denormalization), many times you are able to keep the classes as they are by changing the mapping.
LINQ-to-SQL does nothing of the sort. Your LINQ queries will be tightly coupled to the database structure. If you change the DB, you will probably have to change the LINQ as well.
LINQ to SQL (part of Visual Studio 2008) is an OR Mapper.
LINQ is a new query language that can be used to query many different types of sources.
LINQ itself is not a ORM. LINQ is the language features and methods that exist in allowing you to query objects like SQL.
"LINQ to SQL" is a provider that allows us to use LINQ against SQL strongly-typed objects.
I think a good test to ascertain whether a platform or code block displays the characteristics of an O/R-M is simply:
With his solution hat on, does the developer(s) (or his/her code generator) have any direct, unabstracted knowledge of what's inside the database?
With this criterion, the answer for differing LINQ implementations can be
Yes, knowledge of the database schema is entirely contained within the roll-your-own, LINQ utilizing O/R-M code layerorNo, knowledge of the database schema is scattered throughout the application
Further, I'd extend this characterization to three simple levels of O/R-M.
1. Abandonment.
It's a small app w/ a couple of developers and the object/data model isn't that complex and doesn't change very often. The small dev team can stay on top of it.
2. Roll your own in the data access layer.
With some managable refactoring in a data access layer, the desired O/R-M functionality can be effected in an intermediate layer by the relatively small dev team. Enough to keep the entire team on the same page.
3. Enterprise-level O/R-M specification defining/overhead introducing tools.
At some level of complexity, the need to keep all devs on the same page just swamps any overhead introduced by the formality. No need to reinvent the wheel at this level of complexity. N-hibernate or the (rough) V1.0 Entity Framework are examples of this scale.
For a richer classification, from which I borrowed and simplified, see Ted Neward's classic post at
http://blogs.tedneward.com/2006/06/26/The+Vietnam+Of+Computer+Science.aspx
where he classifies O/R-M treatments (or abdications) as
1. Abandonment. Developers simply give up on objects entirely, and return to a programming model that doesn't create the object/relational impedance mismatch. While distasteful, in certain scenarios an object-oriented approach creates more overhead than it saves, and the ROI simply isn't there to justify the cost of creating a rich domain model. ([Fowler] talks about this to some depth.) This eliminates the problem quite neatly, because if there are no objects, there is no impedance mismatch.
2. Wholehearted acceptance. Developers simply give up on relational storage entirely, and use a storage model that fits the way their languages of choice look at the world. Object-storage systems, such as the db4o project, solve the problem neatly by storing objects directly to disk, eliminating many (but not all) of the aforementioned issues; there is no "second schema", for example, because the only schema used is that of the object definitions themselves. While many DBAs will faint dead away at the thought, in an increasingly service-oriented world, which eschews the idea of direct data access but instead requires all access go through the service gateway thus encapsulating the storage mechanism away from prying eyes, it becomes entirely feasible to imagine developers storing data in a form that's much easier for them to use, rather than DBAs.
3. Manual mapping. Developers simply accept that it's not such a hard problem to solve manually after all, and write straight relational-access code to return relations to the language, access the tuples, and populate objects as necessary. In many cases, this code might even be automatically generated by a tool examining database metadata, eliminating some of the principal criticism of this approach (that being, "It's too much code to write and maintain").
4. Acceptance of O/R-M limitations. Developers simply accept that there is no way to efficiently and easily close the loop on the O/R mismatch, and use an O/R-M to solve 80% (or 50% or 95%, or whatever percentage seems appropriate) of the problem and make use of SQL and relational-based access (such as "raw" JDBC or ADO.NET) to carry them past those areas where an O/R-M would create problems. Doing so carries its own fair share of risks, however, as developers using an O/R-M must be aware of any caching the O/R-M solution does within it, because the "raw" relational access will clearly not be able to take advantage of that caching layer.
5. Integration of relational concepts into the languages. Developers simply accept that this is a problem that should be solved by the language, not by a library or framework. For the last decade or more, the emphasis on solutions to the O/R problem have focused on trying to bring objects closer to the database, so that developers can focus exclusively on programming in a single paradigm (that paradigm being, of course, objects). Over the last several years, however, interest in "scripting" languages with far stronger set and list support, like Ruby, has sparked the idea that perhaps another solution is appropriate: bring relational concepts (which, at heart, are set-based) into mainstream programming languages, making it easier to bridge the gap between "sets" and "objects". Work in this space has thus far been limited, constrained mostly to research projects and/or "fringe" languages, but several interesting efforts are gaining visibility within the community, such as functional/object hybrid languages like Scala or F#, as well as direct integration into traditional O-O languages, such as the LINQ project from Microsoft for C# and Visual Basic. One such effort that failed, unfortunately, was the SQL/J strategy; even there, the approach was limited, not seeking to incorporate sets into Java, but simply allow for embedded SQL calls to be preprocessed and translated into JDBC code by a translator.
6. Integration of relational concepts into frameworks. Developers simply accept that this problem is solvable, but only with a change of perspective. Instead of relying on language or library designers to solve this problem, developers take a different view of "objects" that is more relational in nature, building domain frameworks that are more directly built around relational constructs. For example, instead of creating a Person class that holds its instance data directly in fields inside the object, developers create a Person class that holds its instance data in a RowSet (Java) or DataSet (C#) instance, which can be assembled with other RowSets/DataSets into an easy-to-ship block of data for update against the database, or unpacked from the database into the individual objects.
Linq To SQL using the dbml designer yes, otherwise Linq is just a set of extension methods for Enumerables.