Feature/uhighlighter #2052

idodeclare · 2018-03-28T16:15:08Z

Hello,

Please consider for integration this patch that integrates the Lucene UnifiedHighlighter into OpenGrok's Context [see getContext2()] while preserving for now the original getContext() for falling back when UnifiedHighlighter cannot be used (e.g. when source text has been updated via git-update but Lucene Document postings are not yet updated).

more.jsp and search.jsp both use of getContext2() so their presentations are fully consistent.

OpenGrok's Lucene indexing settings are modified (see OGKTextField) to enable storage of "postings" (i.e., IndexOptions.DOCS_AND_FREQS_AND_POSITIONS_AND_OFFSETS) for use by UnifiedHighlighter so that documents do not have to be re-lexed (as done by PlainLineTokenizer) in order to be highlighted. My testing with some large repos showed only a tiny increase in index size because of this change.

Additionally, for DEFS, tokens must now be indexed with stored term vectors (see OGKTextVecField) or else my testing showed that DEFS multi-term queries (MTQs e.g., prefixes or wildcards) did not operate correctly. This is due to Lucene UnifiedHighlighter currently falling back to ANALYSIS in the presence of MTQs when term vectors are not available. For DEFS, re-analysis from source text does not accord whatsoever with ctags Definitions; but the presence of term vectors makes this work perfectly. (See my long comment in OGKUnifiedHighlighter regarding this issue.)

Since DEFS are a small subset of index data relative to FULL and REFS fields, my testing showed again only a tiny increase in index size due to storage of DEFS term vectors.

This patch also enables optionally the display of additional surrounding lines for each match (as requested in issues #673 and #638). Configuration <contextSurround/> (default 0) specifies the number of lines of leading and trailing context to display surrounding each match.

A new Configuration <contextLimit/> (default 10) specifies the total number of context lines to display for each matching file.

Please see the following new classes that collaborate to produce OpenGrok's style of context under UnifiedHighlighter's coordination:

StrictLineBreakIterator : represents a subclass of Java BreakIterator that breaks at standard OpenGrok EOL -- namely \r\n, \n, or \r, which is needed since OpenGrok's context is always w.r.t. whole source-code lines (which may be truncated to fit within the context "window").
PhraseHighlight: represents a highlighted phrase within a line.
LineHighlight: represents a collection of metadata related to highlighting a single line of code.
PassageConverter: represents an object that can translate Lucene Passage instances into indexed LineHighlight instances, taking into account a configurable number of leading and trailing lines of context for each match.
ContextFormatter: represents a subclass of Lucene's PassageFormatter that uses OpenGrok's PassageConverter.
FormattedLines: represents structured results from ContextFormatter that can be merged with other instances. UnifiedHighlighter returns results separated by field, and OGKUnifiedHighlighter merges them together to return a coherent result for OpenGrok presentation.
OGKUnifiedHighlighter: represents a subclass of Lucene UnifiedHighlighter with customizations for OpenGrok.

There are new tests of this functionality. At the highest level, new SearchAndContextFormatterTest tests the indexing, search, and format of the results. (There are many tests at lower levels.)

Thank you.

Also: - Add Context toggleAlt() so the side effect previously implicit in getContext() is now an explicit, separate call by clients.

tarzanek · 2018-03-29T09:23:09Z

src/org/opensolaris/opengrok/analysis/AnalyzerGuru.java

@@ -531,6 +531,9 @@ public static void writeXref(FileAnalyzerFactory factory, Reader in,
        args.setProject(project);

        FileAnalyzer analyzer = factory.getAnalyzer();
+        RuntimeEnvironment env = RuntimeEnvironment.getInstance();


this is because of testing?

While testing the new highlighter, I found that scopes were not working when looking at historical versions of files or when annotating a file — i.e. where this writeXref() is called by list.jsp instead of its using the stored .gz version.

tarzanek · 2018-03-29T09:52:25Z

what is the difference between CDDL and http://creativecommons.org/licenses/by-sa/3.0/us/ ?
(is CC similar like BSD?)

tulinkry · 2018-03-29T11:47:25Z

src/org/opensolaris/opengrok/analysis/plain/PlainAnalyzer.java

@@ -50,6 +53,9 @@
 */
 public class PlainAnalyzer extends TextAnalyzer {

+    private static final Logger LOGGER = LoggerFactory.getLogger(


yes, removed

ahornace · 2018-03-29T13:12:46Z

src/org/opensolaris/opengrok/analysis/OGKTextField.java

+
+    /**
+     * Creates a new un-stored instance with Reader value.
+     * @param name


Could you please add comments for the parameters? I know that they are mostly self explaining so you could remove them if you prefer to (that's what I would do). It's just that if they have this format then they do not provide any additional info than the parameters already in code. :)

ahornace · 2018-03-29T13:17:23Z

src/org/opensolaris/opengrok/analysis/plain/PlainAnalyzer.java

@@ -134,4 +140,51 @@ public void analyze(Document doc, StreamSource src, Writer xrefOut)
            }
        }
    }
+
+    private void tryAddingDefs(Document doc, Definitions defs, StreamSource src,
+        String fullpath) throws IOException {


Could you please add 4 additional spaces? The same you do in other places :)

There is no ambiguity here; the method body is separated from this with a line break.

ahornace · 2018-03-29T13:24:29Z

src/org/opensolaris/opengrok/analysis/plain/PlainAnalyzer.java

+        return ExpandTabsReader.wrap(reader, project);
+    }
+
+    private static boolean bytesEqual(byte[] a, byte[] b) {


Maybe unused? :) If it is used then you can consider using Arrays.equals().

tarzanek · 2018-03-29T13:50:20Z

src/org/opensolaris/opengrok/index/IndexDatabase.java

+             * should not have a detrimental impact on overall throughput.
+             */
+            iwc.setCodec(new Lucene70Codec(
+                Lucene50StoredFieldsFormat.Mode.BEST_COMPRESSION));


lucene usually has some "latest" links pointing to latest available codecs and formats, does it have it for those two too?
the idea I have is that upon every lucene upgrade everything (possible) should get upgraded automatically to latest greatest
(I know, downside is that you need to index from scratch and all the bugs get automatically picked up, but engine upgrade is transparent)

I did not find any codec class or stored field format enum with "latest" in its name.

ok, we shall stick with this exact matching then, if I will find a way around I will change it to generic to be able to stick to latest lucene (yes, it breaks backwards compatibility, but this is a developer tool, latest greatest is of the essence)

tarzanek · 2018-03-29T13:52:43Z

src/org/opensolaris/opengrok/search/Results.java

            out.write("</td></tr>");
-            for (Document doc : entry.getValue()) {
+            for (int docId : entry.getValue()) {


nice catch and optimalization :)
thank you

tarzanek · 2018-03-29T13:54:36Z

src/org/opensolaris/opengrok/search/Results.java

+            /**
+             * Fall back to the old view, which re-analyzes text using
+             * PlainLinetokenizer. E.g., when source code is updated (thus
+             * affecting SHA hashes) but re-indexing is not yet complete.


old comment? do you use SHA hashes?

ahornace

Nice work! :) It's quite a lot to take all at once. I will try to look at it in more detail later.

ahornace · 2018-03-29T13:28:14Z

src/org/opensolaris/opengrok/configuration/Configuration.java

@@ -389,6 +398,8 @@ public Configuration() {
        setCachePages(5);
        setCommandTimeout(600); // 10 minutes
        setCompressXref(true);
+        setContextLimit((short)10);


Maybe consider using a constant? It might be easier to change later. :)

ahornace · 2018-03-29T13:33:27Z

src/org/opensolaris/opengrok/configuration/Configuration.java

+     * value.
+     */
+    private static final String NONPOSITIVE_NUMBER_ERROR =
+        "Invalid value for \"%s\" - \"%s\". Expected value greater than 0";


I am not very skilled in Java's format but shouldn't the second argument be %d? Since short is being passed there. :)

I agree with the original author that %s is more forgiving for error strings. E.g., if the field were changed to a boxed primitive, the error string would still work (printing null if necessary) but %d might throw an exception.

ahornace · 2018-03-29T13:39:59Z

src/org/opensolaris/opengrok/configuration/RuntimeEnvironment.java

+    public void setContextLimit(Short value)
+            throws IllegalArgumentException {
+        if (value < 1) {
+            throw new IllegalArgumentException("`value' is not positive");


Maybe use the same apostrophe type? :) Now there is ` and '.

That is a quoting style I use for object names to distinguish from e.g. character values, '\0', or from other quoted words.

I would too prefer to use the same character for quoting.

what if we don't use any quotes there? ;) (grokking for that sentence on where it comes from becomes easier)

@tarzanek, ok removed in this patch

ahornace · 2018-03-29T13:51:42Z

src/org/opensolaris/opengrok/search/QueryBuilder.java

+        } catch (ParseException ex) {
+            return fields;
+        }
+        String queryString = query.toString("");


Maybe use query.toString()? :) It translates to query.toString("") in the inside and is nicer.

I prefer to use the explicit Lucene API instead of implicitly using the Java override of toString() that itself uses the explicit Lucene API. Moreover, nothing about the documentation of toString() indicates how it calls the Lucene API; you had to look at source or decompile lucene-core I expect.

ahornace · 2018-03-29T14:03:58Z

src/org/opensolaris/opengrok/search/Results.java

@@ -263,6 +254,46 @@ public static void prettyPrint(Writer out, SearchHelper sh, int start,
        }
    }

+    private static void printPlain(SearchHelper searcHelper,
+        RuntimeEnvironment env, Document doc, int docId, Writer out,


Could you please consider somehow combining the method parameters? This violates current checkstyle parameter number setting which is set to default value, right now it is 7.

ok revised to use a private container object for the arguments to printPlain() that do not change in the calling loop.

ahornace · 2018-03-29T14:10:54Z

src/org/opensolaris/opengrok/search/context/Context.java

+     * @param morePrefix optional link to more... page
+     * @param limit a value indicating if the number of matching lines should be
+     * limited. N.b. unlike
+     * {@link #getContext(java.io.Reader, java.io.Writer, java.lang.String, java.lang.String, java.lang.String, org.opensolaris.opengrok.analysis.Definitions, boolean, boolean, java.util.List, org.opensolaris.opengrok.analysis.Scopes)},


Maybe break the line so it won't be that long? I do not think that it will break the javadoc.

That is auto-generated JavaDoc, so I prefer not to insert a break into it. I do keep it on its own line unchanged (except for a single comma appended for readability) and break the surrounding text.

ahornace · 2018-03-29T14:15:40Z

src/org/opensolaris/opengrok/search/context/Context.java

+        try {
+            List<String> fieldList = qbuilder.getContextFields();
+            String[] fields = new String[fieldList.size()];
+            fieldList.toArray(fields);


Could you please consider using this code: String[] fields = fieldList.toArray(new String[fieldList.size()]); ? I think this might be a little dangerous since it relies heavily on the size of the fields array. If someone would change it then it would break.

ahornace · 2018-03-29T14:28:15Z

src/org/opensolaris/opengrok/search/context/ContextFormatter.java

+            bld.append(HtmlConsts.BR);
+            bld.append("\n");
+            res.setFooter(bld.toString());
+            bld.setLength(0);


bld is local variable so it is eligible for garbage collection so this line is redundant, right? :)

ahornace · 2018-03-29T14:32:59Z

src/org/opensolaris/opengrok/search/context/FormattedLines.java

+     * @throws NoSuchElementException if the instance is empty
+     */
+    public String pop() {
+        String value = lines.remove(lines.lastKey());


Return could be simplified: return lines.remove(lines.lastKey()); :)

ahornace · 2018-03-29T14:35:36Z

src/org/opensolaris/opengrok/search/context/FormattedLines.java

+        for (String line : lines.values()) {
+            bld.append(line);
+        }
+        String f = footer;


Maybe remove the variable f? I think it is nicer to have it just as:

if (footer != null && limited) { bld.append(footer); }

Copying to a local is a habitual style I use when a mutable field is checked for null in order to make impossible a race condition intra-method w.r.t. the field value.

if there are concerns about race condition could it be solved otherwise?
e.g. sync on setter?
I was under the impression this object would be thread local when used ...

Well my point is that my making a local copy is rationalized — i.e. it's a pattern I do recommend generally when nullably-mutable fields are involved so that you can have objects that do not use heavyweight synchronization but also that are not inconsistent intra-method.

A local copy is also valuable in some synchronized settings — e.g., where you have field data that strictly evolves lazily from null to non-null (as in e.g. a singleton) and a volatile field is properly used. In that case too, it's better to make a local copy of the field so that for the predominant case happening after initialization (i.e., when non-null), only a single atomic read is required w.r.t. to field.

ok, makes sense

idodeclare · 2018-03-29T16:16:26Z

what is the difference between CDDL and http://creativecommons.org/licenses/by-sa/3.0/us/ ?
(is CC similar like BSD?)

Yes, CC is a permissive license like BSD which is compatible for use in a CDDL project because CDDL licenses each file individually so that its files can be kept separate from other files with their licenses (permissive or even proprietary) — and neither CDDL nor CC (nor BSD, Apache etc.) restricts the license of the executable version (~~as does~~ unlike GPL).

CC was applicable to those two new classes, OGKTextField and OGKTextVecField, which I imported in a slightly-transformed way from the original source. CC requires attribution, a license link, an indication of any revision, and redistribution of source must be under CC — but otherwise there are no restrictions.

tarzanek · 2018-04-06T08:39:33Z

src/org/opensolaris/opengrok/index/IndexAnalysisSettingsAccessor.java

-        } catch (ClassNotFoundException ex) {
-            // This is not expected, so translate to RuntimeException.
-            throw new RuntimeException(ex);
+        int nres = top.totalHits > n ? n : (int)top.totalHits;


thank you for this update, it will make our life easier with new releases of lucene, where they prefer this new approach to fetching results

tarzanek · 2018-04-06T09:37:30Z

src/org/opensolaris/opengrok/search/context/LineHighlight.java

+    /**
+     * Sets the right elide value.
+     */
+    public void setRelide(int value) {


I am wondering if we should move the project towards scala or kotlin to make the data objects less verbose

come on, less typing ;)
(but then one needs to get ready for some code readability changes - for good or for bad)

tarzanek · 2018-04-06T09:46:56Z

src/org/opensolaris/opengrok/search/context/OGKUnifiedHighlighter.java

+             *     REFS should be approximately fine with re-analysis using an
+             * on-the-fly PlainAnalyzer. It might not accord with the true
+             * language symbol tokenizer, but it should not be wildly
+             * divergent.


can we inject the language symbol tokenizer somehow?

(fwiw we already do that when doing analysis, it's a bit hacky since lucene defaulted to just one default analyzer ( see https://github.com/oracle/opengrok/blob/master/src/org/opensolaris/opengrok/analysis/plain/PlainAnalyzer.java#L98 ) )

OK after studying the Lucene UnifiedHighlighter implementation, I pushed an arcane override of getIndexAnalyzer() when OpenGrok's own highlightFieldsUnion() is executing so that an OpenGrok TYPE-specific analyzer is returned when re-ANALYSIS is being done. (I also updated those block comments.)

tarzanek · 2018-04-06T09:54:53Z

src/org/opensolaris/opengrok/web/HtmlConsts.java

+
+    public static final String BR = "<br/>";
+
+    public static final String HELLIP = "&hellip;";


tarzanek · 2018-04-06T09:57:06Z

src/org/opensolaris/opengrok/web/SearchHelper.java

@@ -520,7 +536,7 @@ public SearchHelper prepareSummary() {
            return this;
        }
        try {
-            sourceContext = new Context(query, builder.getQueries());
+            sourceContext = new Context(query, builder);


especially curious if regexp queries will work

tarzanek · 2018-04-08T13:25:23Z

thank you very much @idodeclare , i don't have any other comments or musings now
so I'd say +1
@vladak @tulinkry @Orviss is it OK to merge?

tarzanek · 2018-04-09T12:06:56Z

@idodeclare could you please ping me a mail to tarzanek at gmail.com with your email address? Since @Orviss will be updating the build we will start using Idea for development, folks from Jetbrains gave us licenses for ultimate, so if you are interested you can have one license ... (the plan is to have the maven used from Idea for development)

tulinkry

copyright issues

tulinkry · 2018-04-11T07:23:50Z

opengrok-indexer/pom.xml

@@ -19,7 +19,7 @@ information: Portions Copyright [yyyy] [name of copyright owner]
 CDDL HEADER END

 Copyright (c) 2010, 2017, Oracle and/or its affiliates. All rights reserved.


tulinkry · 2018-04-11T07:24:36Z

pom.xml

@@ -19,6 +19,7 @@ information: Portions Copyright [yyyy] [name of copyright owner]
 CDDL HEADER END

 Copyright (c) 2010, 2017, Oracle and/or its affiliates. All rights reserved.


tulinkry · 2018-04-11T07:24:45Z

src/org/opensolaris/opengrok/analysis/AnalyzerGuru.java

@@ -19,7 +19,7 @@

 /*
 * Copyright (c) 2005, 2017, Oracle and/or its affiliates. All rights reserved.


tulinkry · 2018-04-11T07:25:32Z

src/org/opensolaris/opengrok/analysis/CtagsReader.java

@@ -378,8 +378,8 @@ private static String cutPattern(String tagLine, int startTab, int endTab) {
     * Adds a tag to a {@code Definitions} instance.


tulinkry · 2018-04-11T07:26:12Z

src/org/opensolaris/opengrok/analysis/archive/TarAnalyzer.java

@@ -19,20 +19,22 @@

 /*
 * Copyright (c) 2005, 2013, Oracle and/or its affiliates. All rights reserved.


tulinkry · 2018-04-11T07:32:47Z

src/org/opensolaris/opengrok/web/SearchHelper.java

@@ -20,14 +20,15 @@
 /*
 * Copyright (c) 2011, 2017, Oracle and/or its affiliates. All rights reserved.


tulinkry · 2018-04-11T07:33:45Z

test/org/opensolaris/opengrok/search/context/ContextTest.java

@@ -83,32 +83,32 @@ public void testIsEmpty() throws ParseException {



old copyright

tulinkry · 2018-04-11T07:34:06Z

test/org/opensolaris/opengrok/search/context/SearchAndContextFormatterTest.java

+ */
+
+/*
+ * Copyright (c) 2008, 2017, Oracle and/or its affiliates. All rights reserved.


wrong copyright

tulinkry · 2018-04-11T07:34:20Z

test/org/opensolaris/opengrok/search/context/SearchAndContextFormatterTest2.java

+ */
+
+/*
+ * Copyright (c) 2008, 2017, Oracle and/or its affiliates. All rights reserved.


wrong copyright

tulinkry · 2018-04-11T07:34:40Z

web/more.jsp

@@ -19,8 +19,8 @@ information: Portions Copyright [yyyy] [name of copyright owner]
 CDDL HEADER END

 Copyright (c) 2010, 2017, Oracle and/or its affiliates. All rights reserved.


idodeclare · 2018-04-11T14:06:32Z

copyright issues

Hi @tulinkry , I do not work for Oracle, so if I were to modify Oracle's notices, it would violate the following part of CDDL section 3.3:

You may
not remove or alter any copyright, patent or trademark notices
contained within the Covered Software, or any notices of licensing
or any descriptive text giving attribution to any Contributor or
the Initial Developer.

The proscription about altering copyright is also in effect when a CDDL-licensed file is copied and altered, as I did with SearchAndContextFormatterTest.java from SearchEngineTest.java (with a JavaDoc comment "Derived from Trond Norbye's SearchEngineTest"). The copyright notices that are not my own are identical between those files.

All the files altered or created by me in this patch include a notice satisfying another part of CDDL section 3.3:

You must include a notice in each of Your Modifications that
identifies You as the Contributor of the Modification

I don't update that manually. I use a script I have, update_cddl.pl, as in e.g.:

    $ git diff master --name-only | xargs update_cddl.pl -P

(-P to use "Portions Copyright" language as appropriate.)

If you like I can post it for use by others. Oracle contributors would run as e.g.:

    $ git diff master --name-only | \
        CDDL_ORG_IDENTITY="Oracle and/or its affiliates" xargs update_cddl.pl -P -r

…y commit)

tarzanek · 2018-04-13T09:54:42Z

thank you Chris, merging, let's play with this and file issues as we go

idodeclare added 4 commits March 28, 2018 08:47

Fix indentation

dc1f06d

Use UnifiedHighlighter subclass for Results, more/ contexts

cee4f4b

Also: - Add Context toggleAlt() so the side effect previously implicit in getContext() is now an explicit, separate call by clients.

Add lucene-highlighter, -join, -memory, -queries Maven dependencies

0299007

Fix required notice

431a6d7

tarzanek added this to the 1.1 milestone Mar 29, 2018

tarzanek reviewed Mar 29, 2018

View reviewed changes

tulinkry reviewed Mar 29, 2018

View reviewed changes

ahornace reviewed Mar 29, 2018

View reviewed changes

tarzanek reviewed Mar 29, 2018

View reviewed changes

ahornace reviewed Mar 29, 2018

View reviewed changes

Fixes from review

efe6abf

tarzanek reviewed Apr 6, 2018

View reviewed changes

Use a TYPE-specific analyzer if re-ANALYSIS is done

f812a10

Revise from review

30f68de

tulinkry mentioned this pull request Apr 10, 2018

normalize fortran identifiers like gfortran #1957

Open

tulinkry reviewed Apr 11, 2018

View reviewed changes

Fix oracle#294. Fix oracle#443. Fix oracle#638. Fix oracle#673. (empt…

fa8513d

…y commit)

tarzanek merged commit bfaf601 into oracle:master Apr 13, 2018

idodeclare deleted the feature/uhighlighter branch April 22, 2018 22:16


		public static final String BR = "<br/>";

		public static final String HELLIP = "…";

		@@ -19,7 +19,7 @@ information: Portions Copyright [yyyy] [name of copyright owner]
		CDDL HEADER END

		Copyright (c) 2010, 2017, Oracle and/or its affiliates. All rights reserved.

		@@ -19,6 +19,7 @@ information: Portions Copyright [yyyy] [name of copyright owner]
		CDDL HEADER END

		Copyright (c) 2010, 2017, Oracle and/or its affiliates. All rights reserved.

		@@ -19,7 +19,7 @@

		/*
		* Copyright (c) 2005, 2017, Oracle and/or its affiliates. All rights reserved.

		@@ -378,8 +378,8 @@ private static String cutPattern(String tagLine, int startTab, int endTab) {
		* Adds a tag to a {@code Definitions} instance.

		@@ -19,20 +19,22 @@

		/*
		* Copyright (c) 2005, 2013, Oracle and/or its affiliates. All rights reserved.

		@@ -20,14 +20,15 @@
		/*
		* Copyright (c) 2011, 2017, Oracle and/or its affiliates. All rights reserved.

		@@ -83,32 +83,32 @@ public void testIsEmpty() throws ParseException {

		@@ -19,8 +19,8 @@ information: Portions Copyright [yyyy] [name of copyright owner]
		CDDL HEADER END

		Copyright (c) 2010, 2017, Oracle and/or its affiliates. All rights reserved.

Feature/uhighlighter #2052

Feature/uhighlighter #2052

Conversation

idodeclare commented Mar 28, 2018

Choose a reason for hiding this comment

Choose a reason for hiding this comment

tarzanek commented Mar 29, 2018

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

ahornace left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

tarzanek Apr 6, 2018 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

idodeclare commented Mar 29, 2018 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

tarzanek commented Apr 8, 2018 • edited Loading

tarzanek commented Apr 9, 2018

tulinkry left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

idodeclare commented Apr 11, 2018

tarzanek commented Apr 13, 2018

tarzanek Apr 6, 2018 •

edited

Loading

idodeclare commented Mar 29, 2018 •

edited

Loading

tarzanek commented Apr 8, 2018 •

edited

Loading