Confusion regarding protobufs - java

I have a server that makes frequent calls to microservices (actually AWS Lambda functions written in python) with raw JSON payloads and responses on the order of 5-10 MB. These payloads are gzipped to bring their total size under lambda's 6MB limit.
Currently payloads are serialized to JSON, gzipped, and sent to Lambda. The responses are then gunzipped, and deserialized from JSON back into Java POJOs.
Via profiling we have found that this process of serializing, gzipping, gunzipping, and deserializaing is the majority of our servers CPU usage by a large margin. Looking into ways to make serialization more efficient led me to protobufs.
Switching our serialization from JSON to protobufs would certainly make our (de)serialization more efficient, and might also have the added benefit of eliminating the need to gzip to get payloads under 6MB (network latency is not a concern here).
The POJOs in question look something like this (Java):
public class InputObject {
... 5-10 metadata fields containing primitives or other simple objects ...
List<Slots> slots; // usually around 2000
}
public class Slot {
public double field1; //20ish fields with a single double
public double[] field2; //10ish double arrays of length 5
public double[][] field3; //1 2x2 matrix of doubles
}
This is super easy with JSON, gson.toJson(inputObj) and you're good to go. Protobufs seem like a whole different beast, requiring you to use the generated classes and litter your code with stuff like
Blah blah = Blah.newBuilder()
.setFoo(f)
.setBar(b)
.build()
Additionally, this results in an immutable object which requires more hoop jumping to update. Just seems like a bad bad thing to put all that transport layer dependent code into the business logic.
I have seen some people recommend writing wrappers around the generated classes so that all the protobuffy-ness doesn't leak into the rest of the codebase and that seemed like a good idea. But then I am not sure how I could serialize the top level InputObject in one go.
Maybe protobufs aren't the right tool for the job here, but it seems like the go-to solution for inter-service communication when you start looking into improving efficiency.
Am I missing something?

with your proto you can always serialize in one-go. You have an example in the java tuto online:
https://developers.google.com/protocol-buffers/docs/javatutorial
AddressBook.Builder addressBook = AddressBook.newBuilder();
...
FileOutputStream output = new FileOutputStream(args[0]);
addressBook.build().writeTo(output);
Also what you might want to do, is to serialize your proto into a ByteArray, and then encode it in Base64 to carry it through your wrapper:
String yourPayload = BaseEncoding.base64().encode(blah.toByteArray())
You have additional library that can help you to transform existing JSON into a proto, such as JsonFormat:
https://developers.google.com/protocol-buffers/docs/reference/java/com/google/protobuf/util/JsonFormat
And the usage is straightforward as well:
to serialize as Json:
JsonFormat.printToString(yourProto)
To build from a proto:
JsonFormat.merge(yourJSONstring, yourPrototBuilder);
No need to iterate through each element of the object.
Let me know if this answer your question!

Related

Java String to Json & vice versa

I want to covert a string based protocol to Json, Performance is key
The String based protocol is something like
<START>A12B13C14D15<END>
and json is
{'A':12,'B':13,'C':14,'D':15}
I can regex parse the string, create a map & serialized to a Json, but it seeems lot of work as I need to convert a stream in realtime.
Would it be more efficient if I just do string manipulation to get the Json output? How can I do the conversion efficiently?
JSON serialization performance is likely not a problem. Don't optimize it prematurely. If you roll your own JSON serializer, you need to put some effort into e.g. getting the escapes right. If the performance does become a problem, take a look at Jackson, which is fairly fast.
Java seems to do regex quite fast, so you might be fine with it but beware that it is quite possible to accidentally build a regex that with some inputs starts backtracking heavily and takes several minutes to evaluate. You could use native String methods to parse the string.
If performance is really a concern, do timing tests on different approaches, select right tools, see what takes time and optimize accordingly.
Lots of ways to go about it, on JSON side. Instead of Map, which is not needed, POJO is often most convenient. Following uses Jackson (https://github.com/FasterXML/jackson-databind) library:
final static ObjectMapper MAPPER = new ObjectMapper(); // remember to reuse for good perf
public class ABCD {
public int A, B, C, D;
}
// if you have output stream handy:
ABCD value = new ABCD(...);
OutputStream out = ...;
MAPPER.writeValue(out, value);
// or if not
byte[] raw = MAPPER.writeValueAsBytes(value);
or, if you want to eliminate even more of overhead (which, really, is unlikely to matter here):
JsonGenerator jgen = MAPPER.getFactory().createGenerator(out);
jgen.writeStartObject();
jgen.writeNumberField("A", valueA);
jgen.writeNumberField("B", valueB);
jgen.writeNumberField("C", valueC);
jgen.writeNumberField("D", valueD);
jgen.writeEndObject();
jgen.close();
and that gets to quite to close to optimal performance you'd get with hand-written code.
In my case I used this library to handle with json in a web application.
Don't remember where to find. May this helps:
http://www.findjar.com/class/org/json/JSONArray.html

Can generate .thrift files from existing java/scala interfaces and data types?

Is there an easy way to take existing Java/scala datatypes and API interfaces and produce corresponding .thrift files? Having Thrift generate server data structures is excessively invasive as it has the consequences:
I cannot annotate my data structures (e.g. for, XML, JSON, hibernate persistence, ...)
this pattern conflicts with other serialization frameworks that want to own, or require modification of my source files.
As a result, its seems like thrift forces itself into being the exclusive persistence format for my server -- unless, that is, I create a data-marshalling wrapper around Thrift or the other my persistence formats that deal with these data structures (hibernate, Jackson, scala BeanProperty, ...). However, this defeats the purpose of an automated data-marshalling tool such as thrift and leads straight to the error-prone world of having to maintain identical-but-separate interfaces and data-structures (= waste of talented engineer time and energy).
I'm totally happy with Thrift auto-generating client code. However, I (strongly) feel that I need the freedom to edit the data structures my server deals with in the APIs.
You can use Swift.
To make a long story short; annotate your classes and interfaces (structs and services in Thrift parlance). Then you can either run Swift's client/server code or you can use the swift2thrift generator to produce equivalent IDL and use the Thrift compiler to generate clients (the latter is what I recommend for what you're describing).
Once that is done to create a TProcessor that you can use in a TServlet with normal TProtocol/TTransport objects, do something like this in your servlet's init():
protected void addProcessor(String name, Object svc) {
ThriftCodecManager codecManager = new ThriftCodecManager(
new CompilerThriftCodecFactory(false)
);
List<ThriftEventHandler> eventList = Collections.emptyList();
ThriftServiceProcessor proc = new ThriftServiceProcessor(codecManager, eventList, svc);
this.processors.put(name, proc);
this.multiplex.registerProcessor(name, NiftyProcessorAdapters.processorToTProcessor(proc));
}
The multiplex instance variable in this example is an instance of TMultiplexedProcessor from libthrift.jar.
Then just do this in your doPost():
#Override
protected void doPost(HttpServletRequest request, HttpServletResponse response)
throws ServletException, IOException {
getServletContext().log("entering doPost()");
TTransport inTransport = null;
TTransport outTransport = null;
try {
InputStream in = request.getInputStream();
OutputStream out = response.getOutputStream();
TTransport transport = new TIOStreamTransport(in, out);
inTransport = transport;
outTransport = transport;
TProtocol inProtocol = getInProtocolFactory().getProtocol(inTransport);
TProtocol outProtocol = getOutProtocolFactory().getProtocol(outTransport);
if (multiplex.process(inProtocol, outProtocol)) {
out.flush();
} else {
throw new ServletException("multiplex.process() returned false");
}
} catch (TException te) {
throw new ServletException(te);
} finally {
if (inTransport != null) {
inTransport.close();
}
if (outTransport != null) {
outTransport.close();
}
}
}
FYI - TJSONProtocol doesn't work with the version of Swift prior to version 0.14 so at this time you'll need to build from source if you need to use that.
Also... Swift forces your structs to be marked final... JPA spec says entities can't be final... seems to work ok with Eclipselink anyhow but YMMV
Since you mention Java: For some project of ours I did implement an Xtext based solution which generates source code and Thrift IDL files from a project specific DSL. Since Xtext/Xtend is based on Java, it may at least be worth a look if that solution could fit your needs. However, I have a slight feeling that it may be overkill in your situation.
As a result, its seems like thrift forces itself into being the exclusive persistence format for my server [...]
That's not Thrifts fault, and it could be any format.
However, I (strongly) feel that I need the freedom to edit the data structures my server deals with in the APIs.
I fully agree. Especially in such cases it is recommended to separate serialization from internal data structures, and look at serialization as what it really is: just one way to manipulate data1). If you have more than one serialization format, you will always end up with a lot of stuff being implemented multiple times in a similar way. This is effect is more or less inavoidable, providing us with the opportunity to do it in whatever is the best way for the project.
The simple serialization strategy works fine with one format. It becomes cumbersome with two/three formats, and finally turns into a real PITA with more than three formats. 2) As always, the maintainable and extendable solution comes at the cost of some added complexity. 3)
You see, altough you may pick one of the formats as the "favourite master data format", technically you don't have to.
(1) A surprising percentage of programming beginners tutorials and books fail to bring that point across properly. (2) This was exactly the use case I solved by means of a DSL as mentioned at the beginning. (3) Yes, I know what YAGNI means. And I know how to estimate the expectation value of risks.
No. Thrift is only supposed to be used for messages between client and server, not for persistence inside server: how would you query it, for example?

Forcing devs to explicitly define keys for configuration data

We are working in a project with multiple developers and currently the retrieval of values from a configuration file is somewhat "wild west":
Everybody uses some string to retrieve a value from the Config object
Those keys are spread across multiple classes and packages
Sometimes the are not even declared as constants
Naming of the keys is inconsistent and the config file (.properties) looks messy
I would like to sort that out and force everyone to explicitly define their configuration keys. Ideally in one place to streamline how config keys actually look.
I was thingking of using an Enum as a key and turning my retrieval method into:
getConfigValue(String key)
into something like
getConfigValue(ConfigKey)
NOTE: I am using this approach since the Preferences API seems a bit overkill to me plus I would actually like to have the configuration in a simple file.
What are the cons of this approach?
First off, FWIW, I think it's a good idea. But you did specifically ask what the "cons" are, so:
The biggest "con" is that it ties any class that needs to use configuration data to the ConfigKey class. Adding a config key used to mean adding a string to the code you were working on; now it means adding to the enum and to the code you were working on. This is (marginally) more work.
You're probably not markedly increasing inter-dependence otherwise, since I assume the class that getConfigValue is part of is the one on which you'd define the enum.
The other downside to consolidation is if you have multiple projects on different parts of the same code base. When you develop, you have to deal with delivery dependencies, which can be a PITA.
Say Project A and Project B are scheduled to get released in that order. Suddenly political forces change in the 9th hour and you have to deliver B before A. Do you repackage the config to deal with it? Can your QA cycles deal with repackaging or does it force a reset in their timeline.
Typical release issues, but just one more thing you have to manage.
From your question, it is clear that you intend to write a wrapper class for the raw Java Properties API, with the intention that your wrapper class provides a better API. I think that is a good approach, but I'd like to suggest some things that I think will improve your wrapper API.
My first suggested improvement is that an operation that retrieves a configuration value should take two parameters rather than one, and be implemented as shown in the following pseudocode:
class Configuration {
public String getString(String namespace, String localName) {
return properties.getProperty(namespace + "." + localName);
}
}
You can then encourage each developer to define a string constant value to denote the namespace for whatever class/module/component they are developing. As long as each developer (somehow) chooses a different string constant for their namespace, you will avoid accidental name clashes and promote a somewhat organised collection of property names.
My second suggested improvement is that your wrapper class should provide type-safe access to property values. For example, provide getString(), but also provide methods with names such as getInt(), getBoolean(), getDouble() and getStringList(). The int/boolean/double variants should retrieve the property value as a string, attempt to parse it into the appropriate type, and throw a descriptive error message if that fails. The getStringList() method should retrieve the property value as a string and then split it into a list of strings based on using, say, a comma as a separator. Doing this will provide a consistent way for developers to get a list value.
My third suggested improvement is that your wrapper class should provide some additional methods such as:
int getDurationMilliseconds(String namespace, String localName);
int getDurationSeconds(String namespace, String localName);
int getMemorySizeBytes(String namespace, String localName);
int getMemorySizeKB(String namespace, String localName);
int getMemorySizeMB(String namespace, String localName);
Here are some examples of their intended use:
cacheSize = cfg.getMemorySizeBytes(MY_NAMSPACE, "cache_size");
timeout = cfg.getDurationMilliseconds(MY_NAMSPACE, "cache_timeout");
The getMemorySizeBytes() method should convert string values such as "2048 bytes" or "32MB" into the appropriate number of bytes, and getMemorySizeKB() does something similar but returns the specified size in terms of KB rather than bytes. Likewise, the getDuration<units>() methods should be able to handle string values like "500 milliseconds", "2.5 minutes", "3 hours" and "infinite" (which is converted into, say, -1).
Some people may think that the above suggestions have nothing to do with the question that was asked. Actually, they do, but in a sneaky sort of way. The above suggestions will result in a configuration API that developers will find to be much easier to use than the "raw" Java Properties API. They will use it to obtain that ease-of-use benefit. But using the API will have the side effect of forcing the developers to adopt a namespace convention, which will help to solve the problem that you are interested in addressing.
Or to look at it another way, the main con of the approach described in the question is that it offers a win-lose situation: you win (by imposing a property-naming convention on developers), but developers lose because they swap the familiar Java Properties API for another API that doesn't offer them any benefits. In contrast, the improvements I have suggested are intended to provide a win-win situation.

Java design question: Is this a good design?

I am writing a Java client that communicates with a remote server via HTTP/XML.
Server sends commands to my client in XML format, like this:
<command>
<name>C1</name>
<param>
.....
</param>
</command>
There is about 10 or more different commands (C1, C2, ...), each of them has different set of params.
My client will process the command, then response server with a execute result, looks like this:
<C1>
<code>200</code>
<desc>xxx</desc>
</C1>
I am only familiar with C, but very new to Java and OOP,
so, my question is simple, how to design the following logic gracefully in a OOP way?
1.Convert the XML string to an XML Object
2.Find correspoding executor based on 'name' element of the XML, and parse the params
3.Execute the command along with the params
4.Convert the result to an XML Object
5.Convert the XML Object to an XML string
Is this a good design?
1.Define an abstract base class and many sub-classes for each command, which include the following method:
void parseCommand(MyXMLObject obj);
void execute();
MyXMLObject generateResult();
or just a simple method:
MyXMLObject execute(MyXMLObject obj);
and these fields:
String mCommandName;
int mRetCode;
String mRetDesc;
2.Then define a factory class to return an instance of one of the sub-classes based on the XML Object.
3.The logic part code:
MyXMLObject obj = MyXMLUtil.getXMLObject(XMLString);
MyCommand command = MyCommandFactory.getCommand(obj);
MyXMLObject retObj = command.execute();
String retStr = MyXMLUtil.getString(retObj);
...//network operation
Generally speaking, it is a good design (you probably need more interfaces and such, and there are various refinements).
The greater problem is that this is, in many ways, a reinvention of the wheel. There are numerous frameworks that handle mapping from POJOs (java objects) to structured textual formats (such as XML or JSON). On top of that, there are many general implementations of command frameworks. It might be worth investigating available options before you provide your own implementation.
In principle, yes, that's how things work. But I think you should separate the different concerns.
Basically I'd say you have a service layer that needs to execute commands and return the results. This layer should know nothing about XML, so I'd design the plain Java Object model and then worry about tying it to your XML infrastructure, probably using an existing XML mapping technology like XStream.
As the others said, the design looks pretty good. To make your life easier, you should have a look at Simple XML to parse XML -> Java and to create XML from Java objects.

Refactoring large data object

What are some common strategies for refactoring large "state-only" objects?
I am working on a specific soft-real-time decision support system which does online modeling/simulation of the national airspace. This piece of software consumes a number of live data feeds, and produces a once-per-minute estimate of the "state" of a large number of entities in the airspace. The problem breaks down neatly until we hit what is currently the lowest-level entity.
Our mathematical model estimates/predicts upwards of 50 parameters for a timeline of several hours into the past and future for each of these entities, roughly once per minute. Currently, these records are encoded as a single Java class with a lot of fields (some get collapsed into an ArrayList). Our model is evolving, and the dependencies among the fields are not yet set in stone, so each instance wanders through a convoluted model, accumulating settings as it goes along.
Currently we have something like the following, which uses a builder pattern approach to build up the contents of the record, and enforce what the known dependencies are (as a check against programmer error as evolve the mode.) Once the estimate is done, we convert the below into an immutable form using a .build() type method.
final class OneMinuteEstimate {
enum EstimateState { INFANT, HEADER, INDEPENDENT, ... };
EstimateState state = EstimateState.INFANT;
// "header" stuff
DateTime estimatedAtTime = null;
DateTime stamp = null;
EntityId id = null;
// independent fields
int status1 = -1;
...
// dependent/complex fields...
... goes on for 40+ more fields...
void setHeaderFields(...)
{
if (!EstimateState.INFANT.equals(state)) {
throw new IllegalStateException("Must be in INFANT state to set header");
}
...
}
}
Once a very large number of these estimates are complete, they are assembled into timelines where aggregate patterns/trends are analyzed. We have looked at using an embedded database but have struggled with performance issues; we'd rather get this sorted out in terms of data modeling and then incrementally move portions of the soft-real-time code into an embedded data store.
Once the "time sensitive" pieces of this are done, the products are flushed to flat files and a database.
Problems:
It's a giant class, with way too many fields.
There is very little behavior encoded in the class; it's mostly a holder for data fields.
Maintaining the build() method is extremely cumbersome.
It feels clumsy to manually maintain a "state machine" abstraction merely for the purpose of ensuring that a large number of dependent modeling components are properly populating a data object, but it has saved us a lot of frustration as the model evolves.
There is a lot of duplication, particularly when the records described above are aggregated into very similar "rollups" which amount to rolling sums/averages or other statistical products of the above structure in time series.
While some of the fields could be clumped together, they are all logically "peers" of one another, and any breakdown we've tried has resulted in having behavior/logic artificially split and needing to reach two levels deep in indirection.
Out of the box ideas entertained, but this is something we need to evolve incrementally. Before anyone else says it, I'll note that one could suggest that our mathematical model is insufficiently crisp if the data representation for that model is this hard to get ahold of. Fair point, and we're working that, but I think that's a side-effect of an R&D environment with a lot of contributors, and a lot of concurrent hypotheses in play.
(Not that it matters, but this is implemented in Java. We use HSQLDB or Postgres for output products. We don't use any persistence framework, partly out of a lack of familiarity, partly because we have enough performance trouble with just the database alone and hand-coded storage routines... we're skeptical of moving towards additional abstraction.)
I had much of the same problem you did.
At least I think I did, sounds like I did. Representation was different, but at 10,000 feet, sounds pretty much the same. Crapload of discrete, "arbitrary" variables and a bunch of ad hoc relationships among them (essentially business driven), subject to change at a moment's notice.
You also have another issue, which you sorta mentioned, and that was the performance requirement. Sounds like faster is better, and likely a slow perfect solution would be tossed out for the fast lousy one, simply because the slower one can't meet a baseline performance requirement, no matter how good it is.
To put it simply, what I did was I designed a simple domain specific rule language for my system.
The entire point of the DSL was to implicitly express relationships and package them up in to modules.
Very crude, contrived example:
D = 7
C = A + B
B = A / 5
A = 10
RULE 1: IF (C < 10) ALERT "C is less than 10"
RULE 2: IF (C > 5) ALERT "C is greater than 5"
RULE 3: IF (D > 10) ALERT "D is greater than 10"
MODULE 1: RULE 1
MODULE 2: RULE 3
MODULE 3: RULE 1, RULE 2
First, this is not representative of my syntax.
But you can see from the Modules, that it is 3, simple rules.
The key though, is that it's obvious from this that Rule 1 depends on C, which depends on A and B, and B depends on A. Those relationships are implied.
So, for that module, all of those dependencies "come with it". You can see if I generated code for Module 1 it might look something like:
public void module_1() {
int a = 10;
int b = a / 5;
int c = a + b;
if (c < 10) {
alert("C is less than 10");
}
}
Whereas if I created Module 2, all I would get is:
public void module_2() {
int d = 7;
if (d > 10) {
alert("D is greater than 10.");
}
}
In Module 3 you see the "free" reuse:
public void module_3() {
int a = 10;
int b = a / 5;
int c = a + b;
if (c < 10) {
alert("C is less than 10");
}
if (c > 5) {
alert("C is greater than 5");
}
}
So, even though I have one "soup" of rules, the Modules root the base of the dependencies, and thus filter out the stuff it doesn't care about. Grab a module, shake the tree and keep what's left hanging.
My system used the DSL to generate source code, but you can easily have it create a mini runtime interpreter as well.
Simple topological sorting handled the dependency graph for me.
So, the nice thing about this is that while there was inevitable duplication in the final, generated logic, at least across modules, there wasn't any duplication in the rule base. What you as a developer/knowledge worker maintain is the rule base.
What is also nice is that you can change an equation, and not worry so much about the side effects. For example, if I change do C = A / 2, then, suddenly, B drops out completely. But the rule for IF (C < 10) doesn't change at all.
With a few simple tools, you can show the entire dependency graph, you can find orphaned variables (like B), etc.
By generating source code, it's going to run as fast as you want.
In my case, it was interesting to see a rule drop a single variable and see 500 lines of source code vanish from the resulting module. That's 500 lines I didn't have to crawl through by hand and remove during maintenance and development. All I had to do was change a single rule in my rule base and let "magic" happen.
I was even able to do some simple peephole optimization and eliminate variables.
It's not that hard to do. Your rule language can be XML, or a simple expression parser. No reason to go full boat Yacc or ANTLR on it if you don't want to. I'll put a plug in for S-Expressions, no grammar needed, brain dead parsing.
Spreadsheets also make a great input tool, actually. Just be strict on the formatting. Kind of sucks for merging in SVN (so, Don't Do That), but end users love it.
You may well be able to get away with an actual rule based system. My system wasn't dynamic at runtime, and didn't really need sophisticated goal seeking and inference, so I didn't need the overhead of such a system. But if one works for you out of the box, then happy day.
Oh, and for an implementation note, for those who don't believe you can hit the 64K code limit in a Java method, well I can assure you it can be done :).
Splitting a Large Data Object is very similar to Normalizing a Large Relational Table (first and second normal form). Follow the rules to reach at least second normal form and you may have a good decomposition of the original class.
From experience working also with R&D stuff with soft real-time performance constrains (and sometimes monster fat classes), I would suggest NOT to use OR mappers. In such situations, you'll be better off dealing "touching the metal" and working directly with JDBC result sets. This is my suggestion for apps with soft real-time constrains and massive amounts of data items per package. More importantly, if the number of distinct classes (not class instances, but class definitions) that need to persisted is large, and you also have memory constrains in your specs, you will also want to avoid ORMs like Hibernate.
Going back to your original question:
What you seem to have is a typical problem of 1) mapping multiple data items into a OO model and 2) such multiple data items do not exhibit a good way of grouping or segregation (and any attempt to grouping tends simply not to feel right.) Sometimes the domain model does not lend itself for such aggregation, and coming up with an artificial way of doing so typically ends up in compromises that don't satisfy all design requirements and desires.
To make matters worse, a OO model typically requires/expects you to have all the items present in a class as class' fields. Such a class is typically without behavior, so it is just a struct-like construct, aka data envelope or data shuttle.
But such situations beg the following questions:
Does your application need to read/write all 40, 50+ data items at once, always?
*Must all data items be always present?*
I do not know the specifics of your problem domain, but in general I've found that we rarely ever need to deal with all data items at once. This is where a relational model shines because you don't have to query all rows from a table at once. You only pulls those you need as projections of the table/view in question.
In a situation where we have a potentially large number of data items, but on average the number of data items being passed down the wire is less than the maximum, you'd be better off using a Properties pattern.
Instead of defining a monster envelope class holding all items :
// java pseudocode
class envelope
{
field1, field2, field3... field_n;
...
setFields(m1,m2,m3,...m_n){field1=m1; .... };
...
}
Define a dictionary (based on a map for example):
// java pseudocode
public enum EnvelopeField {field1, field2, field3,... field_n);
interface Envelope //package visible
{
// typical map-based read fields.
Object get(EnvelopeField field);
boolean isEmpty();
// new methods similar to existing ones in java.lang.Map, but
// more semantically aligned with envelopes and fields.
Iterator<EnvelopeField> fields();
boolean hasField(EnvelopeField field);
}
// a "marker" interface
// code that only needs to read envelopes must operate on
// these interfaces.
public interface ReadOnlyEnvelope extends Envelope {}
// the read-write version of envelope, notice that
// it inherits from Envelope, but not from ReadOnlyEnvelope.
// this is done to make it difficult (but not impossible
// unfortunately) to "cast-up" a read only envelope into a
// mutable one.
public interface MutableEnvelope extends Envelope
{
Object put(EnvelopeField field);
// to "cast-down" or "narrow" into a read only version type that
// cannot directly be "cast-up" back into a mutable.
ReadOnlyEnvelope readOnly();
}
// the standard interface for map-based envelopes.
public interface MapBasedEnvelope extends
Map<EnvelopeField,java.lang.Object>
MutableEnvelope
{
}
// package visible, not public
class EnvelopeImpl extends HashMap<EnvelopeField,java.lang.Object>
implements MapBasedEnvelope, ReadOnlyEnvelope
{
// get, put, isEmpty are automatically inherited from HashMap
...
public Iterator<EnvelopeField> fields(){ return this.keySet().iterator(); }
public boolean hasField(EnvelopeField field){ return this.containsKey(field); }
// the typecast is redundant, but it makes the intention obvious in code.
public ReadOnlyEnvelope readOnly(){ return (ReadOnlyEnvelope)this; }
}
public class final EnvelopeFactory
{
static public MapBasedEnvelope new(){ return new EnvelopeImpl(); }
}
No need to set up read-only internal flags. All you need to do is downcast your envelope instances as Envelope instances (that only provide getters).
Code that expects to read should operate on read-only envelopes and code that expects to change fields should operate on mutable envelopes. Creation of the actual instances would be compartmentalized in factories.
That is, you use the compiler to enforce things to be read-only (or allow things to be mutable) by establishing some code conventions, rules governing what interfaces to use where and how.
You can layer your code into sections that need to write separate from code that only needs to read. Once that's done, simple code reviews (or even grep) can identify code that is using the wrong interface.)
Problems:
Non-public Parent Interface:
Envelope is not declared as a public interface to prevent erroneous/malicious code from casting a read-only envelope down to a base envelope and then back to a mutable envelope. The intended flow is from mutable to read-only only - it is not intended to be bi-directional.
The problem here is that extension of Envelope is restricted to the package that contains it. Whether that is a problem will depend on the particular domain and intended usage.
Factories:
The problem is that factories can (and most likely will) be very complex. Again, the nature of the beast.
Validation:
Another problem introduced with this approach is that now you have to worry about code that expects field X to be present. Having the original monster envelope class partially frees you from that worry because, at least syntactically, all fields are there...
... whether the fields are set or not, that was another matter that still remains with this new model I'm proposing.
So if you have client code that expects to see field X, the client code has to throw some type of exception if the field is not present (or to computer or read a sensible default somehow.) In such cases, you will have to
Identify patterns of field presence. Clients that expect field X to be present might be grouped separately (layered apart) from clients that expect some other field to be present.
Associate custom validators (proxies to read-only envelope interfaces) that either throw exceptions or compute default values for missing fields according to some rules (rules provided programmatically, with an interpreter, or with a rules engine.)
Lack of Typing:
This might be debatable, but people used to work with static typing might feel uneasy with losing the benefits of static typing by going to a loosely typied map-based approach. The counter-argument of this is that most of the web works on a loose typing approach, even on the Java side (JSTL, EL.)
Problems aside, the larger the maximum number of possible fields and the lower the average number of fields present at any given time, the most effective wrt performance this approach will be. It adds additional code complexity, but that's the nature of the beast.
That complexity doesn't go away, and either will be present in your class model or in your validation code. Serialization and transferring down the wire is much more efficient, though, specially if you expect massive numbers of individual data transfers.
Hope it helps.
Actually this looks like a frequent problem that game developers face, bloated classes holding numerous variables and methods because of a deep inheritance tree etc.
There's this blog post about how and why to select composition over inheritance, maybe it would help.
One way you may be able to intelligently break up a large data class is to look at patterns of access by client classes. For example, if a set of classes only accesses fields 1-20 and another set of classes only accesses fields 25-30, maybe those groups of fields belong in separate classes.

Categories

Resources