Daily Archives: October 2, 2017

JavaOne – Apache Maven and its Impact on Java 9

Posted on by
Reply

“Apache Maven and its Impact on Java 9”

Speaker: Robert Scholte

For more blog posts from JavaOne, see the table of contents

Up until Java 8, Maven goes fine on top of Java. In Java 9, there is overlap.

Review of Maven

  • Dependency management
  • All about pom.xml
  • All about classpaths
  • Generate project information

Module path mismatch

  • Classpath accepts – output dirs, jar files and zip files (Maven doesn’t support zip files)
  • Module path accepts output directories, module output directories and directories with jars [and jars; keep reading]
  • Options included pointing to diretory containing the jar (but contains other things to), symlins (windows unhappy), create lib directory and copy files (a lot of IO)
  • Java 9 team added specifying direct jar files per Maven’s request

Module names

  • Like problem with group id/artifact id
  • Need to be unique
  • Learned over 30K artifact ids end with a number. This was a problem for automatic module names because many woul have same name creating collision. ex: commons-lang2 and commons-lang3
  • Also, many artifact ids are naming conflicts if ignore group id. Which is problematic for automatic module names when use two jars that collide
  • Made difference between library builders and app builders. Standard that library builders should never refer to automatic modules. If no exports, can assume it is an application. And apps have control of what gets used.
  • Maven gives warning if library refers to automatic module usage. (use Maven 3.+ so warning shows up in color

Dependencies not yet modular

  • If you are dependent on open source library and it is dependent on one that isn’t modular (or the depenency changed the module name), the direct dependency can’t migrate to modules yet
  • Solution: use Automatic-Module-Name to manifest of libraries so not using algorithm. A good bridge

Tips

  • Point of no return – each time you refer to a dependency that is a module, you can’t go back to a previous version that does not
  • Don’t have split packages if can (Maven itself has a split package issue for old compatibility classes)
  • Don’t have classs in the root package
  • Provide your future module name in the manifest
  • Use a unique module name
  • Don’t change folder structure to add module name
  • plexus-java can help plugins build path
  • Classpath isn’t going away. The older the project, the harder it will likely be to migrate to modules

release tag
In addition to source/target 1.9, you can specify <release> to indicate compatiblity
Can use animal-sniffer to determine if code compiled on a newer version of Java in an older JDK mode will ctually run.

Module file
Maven doesn’t generate. Requires human input. Not all modules are dependencies. jdeps can do it because has access to binary files. But need before compile at build time so not compiling twice

My take: Interesting problems and solutions. I feel like he takeaway is “don’t upgrade to Java 9 yet”. The font was small on the screen. Even when there aren’t a lot of words on the screen

JavaOne – Simplified and Fast Fraud Detection

Posted on by
Reply

Simplified and Fast Fraud Detection”

Speaker: Keith Laker

For more blog posts from JavaOne, see the table of contents

Live SQL

  • free online Oracle 12C database
  • Can save scripts
  • Google searchable
  • Each OTN (oracle tech network) users sees own copy of data. Sandboxed
  • Can download data as CSV

https://livesql.oracle.com/apex/livesql/file/index.html

And for this session the live sql URL

Pattern Matching

  • types – regex, sed/awk
  • in SQL – row level regex
  • new: pattern recognition in a stream or rows – aka can match across rows and columns
  • new SQL construct MATCH_RECOGNIZE – ANSII standard; not Oracle specific

Steps

  1. Bucket and order the data
    • This makes the patterns “visible”.
    • Used order by or partition by/order by so queries are deterministic (this does not require the paid Oracle partitioning feature)
  2. Define the pattern
    • Regular expression like pattern
    • Ex: PATTERN (X+ Y+ Z+) where X/Y/Z is a boolean expression. Ex: bal < PREV(bal)
    • Common qualifiers: * + ? {n} {n,} {n,m}
    • Also have extra ? for reluctant qualifiers – helps deal with what to do with overlapping matches
  3. Define measures
    • Define columns in output table
    • pattern navigation options; PREV, NEXT, FIRST, LAST
    • column
    • optional aggregates (COUNT, SUM, AVG, MAX, MIN)
    • special measures: CLASSIFIER() – which component of the pattern applied to this row and MATCH_NUMBER() – how many matches within each partition – both are good for debugging
    • Ex: MEASURES FIRST(x.tstamp) as first_x
  4. Controlling output
    • by default get a column per measure along with the partitioning column (when using one row per match). Get more columns with all rows per match)
    • how many rows back: ONE ROW PER MATCH (default) ALL ROWS PER MATCH or ALL ROWS PER MATCH WITH UNMATCHED ROWS (good for debugging)
    • where to start next search: AFTER MATCH SKIP PAST LAST ROW (default), also options for next row and relating to variables

Demo

  • Find 3 or more small (<2K) money transfers within 30 days. Then find large transfer (?=1M) within 10 days of last small transfer
  • Can do in SQL without pattern matching, but a lot of code.
  • Can do in Java, but. [copying the database…]
  • Showed how to create a table for JSON data – reads into a CLOB and Oracle checks it is valid JSON. Loaded with insert statements because live sql is web based and can’t access underlying file system.
  • Can use dot notation to access SQL fields

Sample pattern matching statement:


SELECT *
FROM transfers_view
MATCH_RECOGNIZE(
 ORDER BY time_id
 MEASURES
 user_id AS user_id,
 amount AS amount
 PATTERN (X{3,} Y)
 DEFINE
 X AS (amount < 2000) AND 
 LAST(time_id) - FIRST(time_id) < 30,
 Y AS (amount >= 1000000) AND 
 time_id - LAST(x.time_id)< 10);

My take: This was a two hour “tutorial” which differs from a hands on lab. We were still able to follow along with a laptop or “large tablet.” I followed along with the demos on my Mac. Which also let me play a bit. It was fun. I’ve always liked SQL :). I like that he uses QR codes for the links/blogs he wants people to go to. They are also linked in the PowerPoint when it becomes available.

It was also interesting blogging on my laptop. On my tablet, I blog in HTML because it is a pain to u se the visual editor on the tablet. A laptop has no such problem. But a laptop battery doesn’t last all day so…