phoedos/pmd
1
1
Fork 1
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Update doc about "How PMD Works"

Browse Source
This commit is contained in:
Andreas Dangel
2017-09-29 10:30:23 +02:00
parent f7d67df442
commit eb6624deb6
1 changed files with 37 additions and 14 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

51
docs/pages/pmd/devdocs/how_pmd_works.md
View File

@ -1,22 +1,45 @@
---
title: PMD How it Works
title: How PMD Works
tags: [customizing]
summary: How PMD Works
last_updated: July 3, 2016
summary: Processing overview of the different steps taken by PMD.
last_updated: September 2017
permalink: pmd_devdocs_how_pmd_works.html
author: Tom Copeland
author: Tom Copeland, Andreas Dangel <andreas.dangel@adangel.org>
---
# How it works
## Overview
PMD checks source code against rules and produces a report. Like this:
The processing starts e.g. with the main class: `net.sourceforge.pmd.PMD`
* Something passes a file name and a RuleSet into PMD.
* PMD hands an InputStream of the source file to a JavaCC-generated parser.
* PMD gets a reference to an Abstract Syntax Tree back from the parser.
* PMD hands the AST off to the symbol table layer which builds scopes, finds declarations, and find usages.
* If any rules need data flow analysis, PMD hands the AST over to the DFA layer for building control flow graphs and data flow nodes.
* Each Rule in the RuleSet gets to traverse the AST and check for problems. The rules can also poke around the symbol table and DFA nodes.
* The Report is now filled with RuleViolations, and those get printed out in XML or HTML or whatever.
{%include note.html content="This is the command line interface. There are many other means, who
PMD can be invoked. E.g. via ant, maven, gradle..." %}
Not much detail here… if you think this document can be improved, please post [here](http://sourceforge.net/p/pmd/discussion/188192) and let us know how. Thanks!
* Parse command line parameters (see net.sourceforge.pmd.cli.PMDParameters)
Also load the incremental analysis cache file
* Load rulesets/rules
* Determine languages (rules of different languages might be mixed in rulesets)
* Determine files (uses the given source directory, filter by the language's file extensions)
* Prepare the renderer
* Sort the files by name
* Check whether we can use the incremental analysis cache (if the rulesets changed, it will be invalid)
* Prepare the SourceCodeProcessor based on the configuration
* Analyze the files. Either single threaded or multi-threaded parallel. This task is encapsulated
in `net.sourceforge.pmd.processor.PMDRunnable`:
* Create input stream
* Call source code processor (`net.sourceforge.pmd.SourceCodeProcessor`):
1. Determine the language
2. Check whether the file is already analyzed and a result is available from the analysis cache
3. Parse the source code. Result is the root AST node.
4. Always run the SymbolFacade visitor. It builds scopes, finds declarations and usages.
5. Run DFA (data flow analysis) visitor (if at least one rule requires it) for building
control flow graphs and data flow nodes.
6. Run TypeResolution visitor (if at least one rule requires it)
7. FUTURE: Run multifile analysis (if at least one rule requires it)
8. Execute the rules:
* First run the rules that opted in for the rule chain mechanism
* Run all the other rules and let them traverse the AST. The rules can use the symbol table,
type resolution information and DFA nodes.
* The rules will report found problems as RuleViolations.
* Render the found violations into the wanted format (XML, text, HTML, ...)
* Store the incremental analysis cache
* Depending on the number of violations found, exit with code 0 or 4.