Some Eclipse Foundation services are deprecated, or will be soon. Please ensure you've read this important communication.

Bug 357677

Summary:

[parser] Incorrect parsing of XML/HTML escape symbols after editing

Product:

[WebTools] WTP Source Editing

Reporter:

Yahor Radtsevich <yahorr>

Component:

wst.xml

Assignee:

Salvador Zalapa <zalapa>

Status:

RESOLVED FIXED

QA Contact:

Nitin Dahyabhai <thatnitind>

Severity:

normal

Priority:

CC:

nsand.dev

Version:

unspecified

Flags:

nsand.dev: review+

Target Milestone:

3.4.2

Hardware:

OS:

Windows 7

Whiteboard:

Attachments:

Description	Flags
Several screenshots to reproduce	none
Intial Patch	none
Second proposed patch	none
Second proposed patch (fixed)	none
Patch (head version)	none

Description Yahor Radtsevich

2011-09-14 15:38:21 EDT

Build Identifier: 

When XML/HTML escape symbols are edited in XML/HTML editor, they may become unrecognized by the XML/HTML parser.

Reproducible: Always

Steps to Reproduce:
1. Create new XML file with the following content:
&amp;
ASSERT: these symbols are highlighted in blue.
2. Open this file with XML editor.
3. Type space after '&' and remove it.
ACTUAL RESULT:
These symbols become black.
EXPECTED RESULT:
Symbols are highlighted in blue.

Comment 1 Yahor Radtsevich

2011-09-14 15:52:57 EDT

Created attachment 203364 [details]
Several screenshots to reproduce

Screenshot demonstrating this bug are attached. I have used HTML Page Designer to illustrating purposes, but the result is the same for any XML/HTML editor.

Comment 2 Salvador Zalapa

2012-06-14 15:00:25 EDT

Since after breaking down the EntityName region, it becomes in a XML_Content region. So, when the heuristics are triggered (in order do not re parse the region), the XMLContentRegion.UpdateRegion() method consider that if the change's lenght == 0 (withspace deleted) it can handle the change by itself, by just updating the region indexes (without any parse action). This initial patch adds a new heuristic to the XML_Content.UpdateRegion() method, if the content's length is between 4 and 10, it could be a potential EntityName region, so it should be parsed. The scenario reported is covered by this proposed patch however i am still facing an issue why i try to join back the following (due to i am getting 2 regions here):

&Aac ute;

I was wondering if this patch is sufficient and adequate to cover this bug? i am still trying to figure out a solution for the other scenario.

Comment 3 Salvador Zalapa

2012-06-14 15:01:39 EDT

Created attachment 217380 [details]
Intial Patch

Comment 4 Salvador Zalapa

2012-06-28 12:46:23 EDT

Attaching a second patch version, this patch covers all the scenarios. As i said in comment#2, the proposed heuristic consists in detect if the length of the region is between 4 and 10, it is a potential EntityName region, so a reparse is triggered. In this new patch, this also hadles the scenario:

&Aac ute;

For this i did add an extra rule to the XMLTokenizer, to handle a entity region decomposed just as one XML_Content instead of two or more (in order to be cached just one region, so the reparse can be performed properly).

Comment 5 Salvador Zalapa

2012-06-28 12:47:05 EDT

Created attachment 218036 [details]
Second proposed patch

Comment 6 Salvador Zalapa

2012-06-28 15:24:04 EDT

Created attachment 218050 [details]
Second proposed patch (fixed)

In the last patch i forgot to delete some system.out.println sentences.

Comment 7 Salvador Zalapa

2012-07-04 12:38:03 EDT

Created attachment 218284 [details]
Patch (head version)

My mistake again, the last patch used an old WTP version, this one is on head version.

Comment 8 Nick Sandonato

2012-11-07 16:37:49 EST

Hi Chava, is there any way to accomplish this without being dependent on the length of the region being between 4 and 10 characters? Maybe something based on the text contents or region type.

Comment 9 Salvador Zalapa

2013-01-07 16:37:29 EST

https://github.com/zalapa/webtools.sourceediting/commit/425850e9d8f296d0c0ea156eba27e251f3c49386

Adding the new version, this is filtering the text started with "#" and ended with ";"

Comment 10 Nick Sandonato

2013-01-08 10:38:48 EST

Patch from the remote repository looks good. Thanks, Chava.