The Weird and The Wonderful

My issue with libraries is that they often do something that's nearly but not quite what you want, and you end up changing what you want to fit the library, which is backwards. This is particularly true with more complex frameworks like Hibernate, Spring etc. A good example from my personal experience is lambdaj[^], a fluent query library in Java. It looks great, until you realise that it uses dynamic proxies so you can't use it on a list of strings, it loads everything into memory so you can't use it on external data sets, and to work around those two things is more effort than writing the subset of query functionality that you need.

On a project level, it also means that anyone new to the team has to learn a new library, whereas presumably they are familiar with the standard library (java.*, System.* etc) which a home-rolled version will use.

If it really, actually, does do the thing that you want, then fair enough. But it's surprising how often you find that a little way down the road you have nasty wrinkles in your code where it butts up against the library, because that third party tool doesn't actually do things the way you want to do them.

I understand what you're getting at. I've also come across libraries that haven't quite done what I was after. In that case you're completely correct.

Obviously I'll evaluate a library first. It's not as if I download the first thing I find and plump for that like it's cool. It also has to have good documentation and positive peer reviews because without them you're on your own, with someone else's code.

You can't guarantee that it will be any easier picking up someone's homebrewed alternative. You'd still have to learn that and it might not have the same documentation or review coverage as the third party component. This is where I find independent peer reviews are worth their weight in gold.

At any rate, I've still got 39 of the 40 estimated hours left for ironing out my wrinkles.

totally agree.

"Program testing can be used to show the presence of bugs, but never to show their absence."

<< please vote!! >>

The make-it-from-scratch vs. use-third-party-solution debate will never end.

Each choice has different, uncomparable costs and benefits, so it's fundamentally an apples vs. oranges debate. It's perfectly normal for the final decision to be made due to irrational or irrelevant reasons.

Most people err too much on the side of make-it-from-scratch, though, as they don't want to go through the effort of figuring out what third party components exists and are good enough for the task. However, I've seen people who dislike coding (but have to do it as part of their job) err on the side of trying to get third party stuff do everything for them.

Anyway, the main reasons to not go for a third party solution are:
1. It costs too much / license is not flexible enough.
2. The problem it solves is too simple; making your own solution takes less effort than figuring out theirs.
3. Whether or not it will be maintained in the future is questionable; maybe you'll have to write your own solution, anyway.
4. It's too buggy or poorly implemented.
5. Your problems are slightly different from the problems it solves.

The biggest challenge is legal. Do the libraries explicitly grant you an IP license, do they specifically state they are free of encumbrances of any sort, do they clearly give you the right to sell and/or distribute the code?

I have seen deals go bad over licensing third party code - both from the point of buying a component to the issue in buying an entire product or company. Be careful.

But don't reinvent the wheel. There are tons of third party libraries with excellent licenses that save a TON.

cheers,
Chris Maunder

The Code Project | Co-founder
Microsoft C++ MVP

I pretty much won't consider third party stuff if it requires another dev to do some non-obvious step to get it running on their computer. If they just have to get latest from TFS and it will work, that's great. If they have to run some license installer on their machine before it will work, that's a no go.

Thou mewling ill-breeding pignut!

I agree with Bobjanova. I'm not against using libraries, but you have to think it through thoroughly and if possible accept it's limitations, something not always easily explained to managers.

Another option is to create your own reusable framework (for very specific stuff).

V.

(MQOTD Rules and previous Solutions )

I'm not against third party stuff in general, but..

jim lahey wrote:
as we're not modifying it and it's already freely available, we won't have a problem redistributing it.

You don't have to modify it .. yet. Someday there will be a problem where you have to choose between modifying the library or some worse option. That's not uncommon (what library really does exactly what you wanted, rather than exactly what someone else wanted? often the problem is hard to work around, too), and it's often bad news for the license.

Some other reasons I frequently have are
- Is crap. No contest there, I think. No one likes a library that's crap.
- Impossible to use. Applies to 80% of all open source libraries pretending to be "portable". What they really mean is "works on one's-complement processors and platforms where ints are 27.5 bits" - it won't compile with MSVC.
- Not as good as I could make it. Sure it'd save me time, but I don't like using bad solutions just because they're easy. Probably doesn't apply to a widely used 5-rated library though.
- Integrates in an invasive or crappy way. Especially happens with libraries that are too framework-like.
- Is a moving target. Not so bad if you can ignore new versions, but sometimes the new version has a feature that would save you a lot of time .. and also has breaking changes. Being forced to make a choice like that pisses me off. Bonus points if that happens at least twice a year. Sometimes old versions even have an expiration date (anyone who does this can go die in a fire), or it auto-updates itself (again, go die in a fire. no one pulls this off without ever introducing a problem - that I will get blamed for of course).
- Is dead. Dead projects don't inspire much confidence in me, even if good things are claimed about them. Bonus point for pretending to be alive when they're not.

When you are developing a long lasting product, using third party components has several risks:

- you can face a bug all of a sudden and get in deep trouble - usually at the worst moment - because there is no maintainer available (the company went out of business, the freeware project died or you can't afford the cost);

- trying to fix a bug yourself from open sources can be a nightmarish endeavour; libraries can be huge and written by gurus in the least readable way; architecture documents are often unavailable; all you can do looks like quick and dirty hacks;

- adding a feature of your own when required is an assault course as well;

- packages are often overfeatured for your needs; you'll have to carry a monster (redistribute it and complexify your installers) for the rest of your life;

- big packages are sometimes so complex that the effort it takes to master the API and comply with it may not be worth;

- last but not least, portability is the main thing; my code needs to be compatible with all MS compilers from VC6 to VC2012, as well as gcc, and future compilers; if a third party library fails to support one of these - which is quite likely to arise -, I can get in big trouble; and this is unpredictable and out of my control.

When dealing with your core business, integrating black boxes is not so reasonable. In the long run, you need full mastership of your source code.

Third-party technology is just great...when it's right at all.

Let's leave aside considerations such as bugs in the package. IBM's old mantra that "you never get the last thousand bugs out" will be true until the Sun goes nova, and we all know it. But the probability of such bugs in a third-party package is much less than that of bugs in one's own, newly designed solutions, so it's a false trail in discussions like this one.

The ugly secret about third-party tech is that packaged solutions are never free of constraints. To use a third-party library, you have to be reasonably certain that the constraints it will impose on your application are bearable. Sometimes, you have to err on the side of caution and decline to take the risk.

My business is hard-real-time, so I'm very sensitive to that. If a third-party package can't guarantee me performance within my time constraints, then I can't use it. Not "won't." Can't.

Other sorts of development have other needs. I recently turned away from a popular XML parser because it imposed a callback requirement on a legacy application whose architecture was inimical to that approach. In another recent case, I spurned a set of .NET facilities because they'd clash with my need to keep the addresses of certain dynamic memory constructs stable. No doubt there are other cases of this sort that I just can't recall this morning.

The application is the ultimate determinant. Does the third-party library you're considering fit within the application's constraint envelope? Does it make "philosophical" demands on the application's architecture? Might there be forward-looking consequences to using it that should be considered in the present? Will your wife find out? Everything matters. As some very bright people have said:

"You can never do only one thing." -- Marc Stiegler.

"TANSTAAFL!" -- Robert A. Heinlein.

(This message is programming you in ways you cannot detect. Be afraid.)

I am very well aware of third party tools. Actually we use one which costs a lot. The project on it has about 20 Forms, 30 UserControls, 10 classes, each one using that software more or less.

The first 2 years worked out fine. Then came a "major upgrade". And, guess what: they changed about all of the namespaces (from productxyz.xxx to productxyz.xxx.zzz). And nothing worked anymore, so rebuild, reinstall over a 100 clients, ecc...

And the next case: well known ChartFx. License (even paid) after 7 years expired.. What dit I do: forget ChartFx and use Microsoft.Net Chart...

I am with you totally. I see complaints like the ones which have already followed your post all the time. Fact is most people have such large egos that it is easier for them to write something than to learn to use code someone else wrote - because they are already "experts" and no longer students.

We use boost all the time for as much of what it can do as we can possibly use. We use Google code (kml libs), we use zlib, we pretty much use anything open source that fits our needs.

If a half decent C++ dev can't wrap a library to give us a stylistically appropriate interface in less time than it would take to write the entire library, then that C++ dev is no use to us.

Most of the arguments against are straw men (presuming open source here). Will a 3rd party library be maintained going forward? Dunno - but sure as heck our own home grown solution won't be maintained by anyone else either. Is it any good? There are a large number of excellent libraries out there for many, many common programming needs. They have had thousands of eyeballs and hundreds of consumers testing them and fixing them. Sure beats the size of our QA team. Code Project itself is an excellent resource for working code that has been peer reviewed the heck out of.

Do your research. Set aside your ego. Borrow code.

I've been a little surprised at just how negatively people perceive using 3rd party stuff.

Your point about ego is spot on.

I think there's a bit of a knee-jerk reaction among most developers when facing a need like that: hell I'm a programmer, this is what I do, I'll write it myself and get exactly what I want.

But that isn't always the right way to go. I used to work at a software company where just about all of the tools were built in-house, from source control to back-end processing frameworks to load balancing. This was a fairly small shop where there weren't really enough programmers for the workload and projects were always behind. It amazed me that so much time was spent building and maintaining software that could be pulled off a shelf, when the programming that really had to get done was behind.

Worse, due to time constraints the in-house software was frighteningly buggy and unreliable, and no one wanted to fix or maintain any of it (especially the guys who wrote it). It got so bad that in one meeting the CEO blew up on the guy who was the main push for in-house solutions, pretty much saying that it had been a mistake. From his point of view, he was paying programmers to write code that was already written and didn't make money, instead of getting cracking on the code that actually did make money. When you're a software company, you don't want your programming resources tied up in creating software you can't sell.

Of course, the main argument for the in-house solutions was that the off-the-shelf solutions didn't fully meet the needs, and that was somewhat true although those things could have been worked around. I guess the moral of the story is, if you want to go the in-house route, make sure that you have the resources to do it right and maintain it. And make sure that management knows that it will require extra time and resources (a.k.a money), and that it will be hard to go back once the in-house software is entrenched.

I'm not dismissing the use of libraries, I just haven't had good experiences with them so far.

We once ran into a bug with a communications library (back in the days of dial-up) and after we identified it, we called the vendor. They said it was a known issue and would be fixed in the next release. We said, "Great, when's the next release?" They said, "In six months." We had a 6 week delivery date. So we had to trash it and write our own, which ended up being smaller and faster.

Libraries add bloat. The library does 10 things, you're only interested in 2 of them.

Management decided to get involved because they didn't like our delivery estimate and so they went out and purchased a library that they had seen a demo of and then gave it to us expecting that all we had to do was integrate it into our existing system. That's simple, right? No excuses for bugs since it was a purchased product the only cause for errors would be from our staff. As you can imagine, it didn't go well and caused much animosity with management that didn't need to be there.

Psychosis at 10
Film at 11

Those who do not remember the past, are doomed to repeat it.
Those who do not remember the past, cannot build upon it.

BrainiacV wrote:
gave it to us expecting that all we had to do was integrate it into our existing system. That's simple, right?

Haha, reminds me of my old CEO. Whenever asking for (sometimes major) code changes he'd always say something like "it's an object, right? Don't you just have to change a couple of lines?"

Apparently someone had explained OO to him in glowing terms at some point. A little knowledge can be a dangerous thing.

Third party drawbacks:
1. Have to learn their api. sometimes its mind bending sometimes simple.
2. Depend on their support/consceintiousness for bug fixes and upgrades.
3. Can't fix a bug if you have a special case. even if you get the source code.
4. Don't always know you cant solve a problem until you have wasted a week on the componenet.
5. Don't own it cant enhance it.

Pluses:
1. debugged already
2. learn from other peoples problems - may get features you hadn't even considered.

Great article.

It is supposed third party tools are the salvation to those activities one is not the best at. But, how many hours are spend before buying? How many hours are dedicated to understand other's logics? What if the documentation lacks of good examples to understand how to get the best of tool?

Sometimes, you have to start from scratch.

CarlosAlbertoEstrada wrote:
How many hours are dedicated to understand other's logics? What if the documentation lacks of good examples to understand how to get the best of tool?

Actually, that can be much worse with an in-house solution, especially once the guy who wrote it is gone. It's all fun and games until there's no one around who knows how it works anymore. At least with a third-party solution, there is likely to be a community of other people who use it and who can answer questions, even if the company that made it no longer maintains it. And...documentation? At least the third-party solution is likely to have some Smile | :)

I use third party libraries (and tools) whenever I consider that the effort of writing my own solution will outweigth the effort of making the third party solution work, usually I evaluate different solutions and pick the best/easy to use/better reviewed/open one; the only times I avoid a third party library is when their license is incompatible with closed source (I work with closed source systems) which invariably means when they use GPL.

CEO at:
- Rafaga Systems
- Para Facturas
- Modern Components
for the moment...

jim lahey wrote:
We want to be able to interpret expressions defined in Xml files

Presumably because this reflects a real business need versus a 'really cool' development idea.

jim lahey wrote:
and we've got a component that can do a hell of a lot more to boot

Which isn't a recommendation. After all Oracle can do a lot more than store a configuration value but one certainly shouldn't choose to use Oracle only to store a configuration value.

jim lahey wrote:

What objections do you have to using third party stuff?

Some Reasons
- Often the complexity greatly exceeds the needed use.
- The intended use is much more suited to a far simpler solution. This is similar to but not the same as the prior reason. It comes about because the developer finds a library they want to use and then invents a problem to solve with it or rationalizes much more needed functionality for an existing problem.
- The license is not compatible with the needs of the business. This often is because some licenses can require that the Application code must be available as source. Another common problem is that that there is no license which means one must have explicit use license from the author to use it.
- It isn't available as source and the author(s) support is erratic.

One must of course have a policy in place to keep and track license(s). And one should keep an actual copy of the license not just a link. One must also be prepared for the possibility that at some time the license might change.

The built in profanity filter should indicate the unfortunate alias selected by a female colleague whose first language is not English! Blush | :O

What a sunshine!

SQL

select
count(usr.usr_id) c***
FROM   [dbo].[user_details]  usr
       LEFT JOIN [dbo].[aspnet_Users]  a_usr ON  usr.usr_id = a_usr.usr_id
       where a_usr.usr_id IS NULL

It's well known that if all the cat videos and porn disappeared from the internet there would be only one site left and it would be called whereareallthecatvideosandporn.com

Brilliant.

Maybe she was having a bad day?

What a count!

The universe is composed of electrons, neutrons, protons and......morons. (ThePhantomUpvoter)

Yeah, it really should have been:

SQL

count(usr.usr_id) AS NumberOfOrphanedCunts

Always strive to make the intent clear.

The real weird thing is however the nullable foreign key.