Use ICU for text search #15858

lhecker · 2023-08-21T12:34:40Z

The ultimate goal of this PR was to use ICU for text search to

Improve Unicode support
Previously we used towlower and only supported BMP glphs.
Improve search performance (10-100x)
This allows us to search for all results in the entire text buffer
at once without having to do so asynchronously.

Unfortunately, this required some significant changes too:

ICU's search facilities operate on text positions which we need to be
mapped back to buffer coordinates. This required the introduction of
CharToColumnMapper to implement sort of a reverse-_charOffsets
mapping. It turns text (character) positions back into coordinates.
Previously search restarted every time you clicked the search button.
It used the current selection as the starting position for the new
search. But since ICU's uregex cannot search backwards we're
required to accumulate all results in a vector first and so we
need to cache that vector in between searches.
We need to know when the cached vector became invalid and so we have
to track any changes made to TextBuffer. The way this commit solves
it is by splitting GetRowByOffset into GetRowByOffset for
const ROW access and GetMutableRowByOffset which increments a
mutation counter on each call. The Search instance can then compare
its cached mutation count against the previous mutation count.

Finally, this commit makes 2 semi-unrelated changes:

URL search now also uses ICU, since it's closely related to regular
text search anyways. This significantly improves performance at
large window sizes.
A few minor issues in UiaTracing were fixed. In particular
2 functions which passed strings as wstring by copy are now
using wstring_view and TraceLoggingCountedWideString.

Related to #6319 and #8000

Validation Steps Performed

Search upward/downward in conhost ✅
Search upward/downward in WT ✅
Searching for any of ß, ẞ, ss or SS matches any of the other ✅
Searching for any of Σ, σ, or ς matches any of the other ✅

lhecker · 2023-08-21T12:37:11Z

src/buffer/out/Row.hpp

+    template<typename T>
+    T _adjustBackward(T column) const noexcept;
+    template<typename T>
+    T _adjustForward(T column) const noexcept;


These were made into templates so that I can use ptrdiff_t indices with them.

src/buffer/out/Row.cpp

lhecker · 2023-08-21T12:58:49Z

src/buffer/out/search.cpp

-    _needle(s_CreateNeedleFromString(str)),
-    _renderData(renderData),
-    _coordAnchor(s_GetInitialAnchor(renderData, direction))
+// - reverse - True when searching backward/upwards in the buffer


After going through all the callers of Search I found that just using bool parameters is more ergonomic. None of the callers, not even conhost, actually used the two enums. They just converted them back/forth.

I would suggest reviewing this file without the diff viewer. It's a mostly new implementation.

lhecker · 2023-08-21T13:03:41Z

src/host/telemetry.cpp

-    _fpDirectionDownAverage = 0;
-    _fpMatchCaseAverage = 0;
-    _uiFindNextClickedTotal = 0;
-}


It seems no one ever noticed that we passed reverse == true (= upwards_search == true) as the fDirectionDown parameter. In other words, the telemetry for DirectionDownAverage was broken. Since I feel like no one is using this telemetry anyways, I chose to remove it instead of fixing it.

lhecker · 2023-08-21T13:17:26Z

src/types/UiaTextRangeBase.cpp

-    auto searchAnchor = _start;
-    if (searchBackward)
+    const std::wstring_view queryText{ text, SysStringLen(text) };
+    const auto results = _pData->GetTextBuffer().SearchText(queryText, ignoreCase != FALSE, _start.y, _end.y + 1);


Previously, this code could use the Search class, because Search would stop searching once it found something. The new Search class will go through the entire buffer however.

To fix that we could either add 2 parameters to Search to specify a bounding range for the search or use TextBuffer's function directly. I felt like the latter is a better option, because while the 2 for loops below are verbose and somewhat difficult to understand initially, they're also easier to verify for their correctness IMO.

@carlos-zamora I didn't quite understand the UiaTextRangeBase::FindText implementation. Could you review that file for me? 🙂 Also, how do I best test that function?

Reviewing 😊

As for how to test it on your own, you can do that using "Accessibility Insights":

open A11y insights

Click on the terminal

Under "Patterns", Click "Explore..." button next to "TextPattern"

A "TextPattern Explorer" window should appear

Click the "..." button next to the "Document" range

Click "Find..."

This new window should allow you to directly interact with the API much like any screen reader could.

It seems Accessibility Insights is kinda buggy? I can't really tell how it works, but if you open the Find dialog and ever reach a "not found" situation, it'll continue to say "not found" forever.

After testing it, I think that my implementation works correctly. If I type "eeeee" the find dialog will find all "e"s no matter where they are or whether I use forwards or backwards search. So there don't seem to be any off-by-1 errors in the code.

However, there's one big difference to the old code: My new implementation returns a "not found" result when it has reached the end of the buffer. The old code continued to circle around the buffer if you reached the end. (This didn't always work for the backward search due to the start > _start check. It's supposed to be start >= _start, otherwise it can't find results that are at the very start of the buffer at 0,0.)
Was that intentional? Should the UIA search circle around in the buffer when it has reached either end?

Previously, this code could use the Search class, because Search would stop searching once it found something. The new Search class will go through the entire buffer however.

Won't this cause a decent performance hit? _getOptimizedBufferSize() used to reduce the search space by ignoring anything past the mutable bottom. Now that we're searching through the whole buffer again, won't FindText('a', ...) be slower. And if the user calls that function repeatedly, the user may have to wait quite a bit before being able to perform the next search? (FYI I haven't verified the performance question via a11y insights yet, so this could all be false)

It seems Accessibility Insights is kinda buggy? I can't really tell how it works, but if you open the Find dialog and ever reach a "not found" situation, it'll continue to say "not found" forever.

You could also use inspect.exe. It's the predecessor to a11y insights and can be found in the Windows SDK's \bin\<version>\<platform> folder. To interact with the API, do the following:

open inspect.exe

click on the terminal (the terminal should be highlighted in the UIA tree in inspect.exe)

Click "Action" > "Text Pattern Explorer..." (a new "Text Explorer" window should appear)

Click "Range" > "Find..."

Was that intentional? Should the UIA search circle around in the buffer when it has reached either end?

Interesting! I was originally going to say that we should keep the same behavior as before, but it looks like this experience is different from MS Word's! There were two notable differences with your implementation:

circling the buffer: MS Word doesn't circle the text area whereas we do in Terminal (old and new impl). So, we probably shouldn't do that anymore.

no more results: MS Word returns something that is interpreted by inspect.exe as "ProcessFind: FindText couldn't find matching text.". But Terminal (new impl) is interpreted as "ProcessFind: MoveEndpointByRange failed. Error -2147024809".

We should align these experiences (which is a bit annoying because your PR wasn't really intended to deal with any of this). If that's easy to do, it'd be best to do that as a part of this PR or in a follow-up before the next release. If not, I think we should keep the old experience, but I know that's not really possible anymore given that the Search class is now completely different. Thoughts?

Previously, this code could use the Search class, because Search would stop searching once it found something. The new Search class will go through the entire buffer however.

Won't this cause a decent performance hit? _getOptimizedBufferSize() used to reduce the search space by ignoring anything past the mutable bottom.

The new ICU-based searcher can search the entire text buffer in 4ms.

It also avoids searching any part of the text buffer that has not been "committed" to memory. :)

ICU also has usearch_previous. We could use it for UIA specifically since it doesn't need the total number of results unlike WT (which is why it uses uregex because it's way way faster when searching for all results). Just like uregex it's very easy to use. I can make it fit. 🙂

And I'll try to make our UIA function behave like Word.

Nevermind, it seems that searching through a character-iterator is a C++ API feature and Windows only ships the C API. Well it's probably fine. I feel like this is not going to be a bottleneck no matter what we do.

zadjii-msft · 2023-08-21T15:30:33Z

URL search now also uses ICU

Does it just so happen to work across wrapped lines now? @asklar literally reported that to me on Teams this weekend 😄

lhecker · 2023-08-21T16:22:16Z

lhecker · 2023-08-21T16:50:47Z

Oh wait, I think you meant a different problem, right? I think that's something like Terminal::GetHyperlinkAtBufferPosition, because that one searches only a single line?

zadjii-msft

leaving notes before lunch

zadjii-msft · 2023-08-21T15:57:08Z

src/buffer/out/textBuffer.hpp

@@ -122,7 +124,7 @@ class TextBuffer final
    ROW& GetScratchpadRow();
    ROW& GetScratchpadRow(const TextAttribute& attributes);
    const ROW& GetRowByOffset(til::CoordType index) const;
-    ROW& GetRowByOffset(til::CoordType index);
+    ROW& GetMutableRowByOffset(til::CoordType index);


seems like a ton of this diff is just this rename

Yeah. But it's a necessary change because many callers have a TextBuffer& but only need a const ROW&. (TextBufferCellIterator for instance is widely used and does this.) 😟

src/buffer/out/Row.cpp

src/cascadia/TerminalControl/ControlCore.cpp

src/buffer/out/textBuffer.cpp

zadjii-msft · 2023-08-21T16:55:01Z

src/buffer/out/search.cpp

+    _results = textBuffer.SearchText(str, caseInsensitive);
+    _mutationCount = textBuffer.GetMutationCount();



presumably, the caller has a write lock on the text buffer so the MutationCount can't change between the search and getting the count, but figured I'd mention it here. If there's a way to guarantee that, it might be worthwhile.

We could only assert that the lock is being held, but I think that this might be better as a separate PR that does this for all the parts that implicitly rely on the console lock being held. The mutation count could be made atomic, but this won't realistically improve anything, because you need to hold the lock anyways to initiate any kind of search (either to do a new search or to select stuff in the buffer).

src/buffer/out/Row.cpp

DHowett · 2023-08-21T17:50:58Z

src/buffer/out/Row.cpp

+    const til::CoordType columnCount = _columnCount;
+    const til::CoordType scale = _lineRendition != LineRendition::SingleWidth;
+    const til::CoordType padding = _doubleBytePadded;
+    return (columnCount - padding) >> scale;


I'm surprised padding comes before un-scaling, but I think it makes sense? A double-byte padded character that doesn't fit on a double-width row didn't fit because it took up 4 columns.. hm.

Yeah, I was really unsure in what order these need to be, and I still am a little.
Thinking about it some more, does it have to be like this?

const auto padding = _doubleBytePadded << scale; return (columnCount - padding) >> scale;

That would be the same as...

return (columnCount >> scale) - padding;

right?

Not quite since the rounding is different. In your code it would subtract 2 no matter what. For instance, imagine padding is 2 and columnCount is 11. In my code snippet it would return 4 and with yours it would return 3. I think 4 is correct because 4*2+3 = 11 (3 being the max. wide glyph padding).

DHowett · 2023-08-21T17:53:00Z

src/buffer/out/Row.hpp

@@ -65,6 +65,28 @@ struct RowCopyTextFromState
    til::CoordType sourceColumnEnd = 0; // OUT
 };

+// This structure is basically an inverse of ROW::_charOffsets. If you have a pointer
+// into a ROW's text this class can tell you what cell that pointer belongs to.
+struct CharToColumnMapper


nobody consumes this externally, do they? Can it be a forward declaration?

I was planning to use this class in other parts of the code base in the future (the renderer primarily).

DHowett · 2023-08-21T17:53:37Z

src/buffer/out/lib/bufferout.vcxproj

@@ -26,6 +26,7 @@
    <ClCompile Include="..\precomp.cpp">
      <PrecompiledHeader>Create</PrecompiledHeader>
    </ClCompile>
+    <ClCompile Include="..\UTextAdapter.cpp" />


update sources! and probably the sources for the UT too.z

src/buffer/out/textBuffer.cpp

DHowett

27/35

DHowett · 2023-08-21T18:02:58Z

src/cascadia/TerminalControl/ControlCore.cpp

+            _searcherGoForward = goForward;
+            _searcherCaseSensitive = caseSensitive;
        }


Assuming you don't want to make these getters on Search?

After thinking about it some more, I'd like to do that a different time if anything. Other editors reset the search when there's any change whatsoever. Technically we shouldn't need to track these members - we should just reset the search whenever any field in the dialog is changed. I just haven't done the work for that.

src/buffer/out/textBuffer.hpp

DHowett · 2023-08-21T18:06:45Z

src/buffer/out/textBuffer.hpp

@@ -373,12 +372,15 @@ class TextBuffer final

    TextAttribute _currentAttributes;
    til::CoordType _firstRow = 0; // indexes top row (not necessarily 0)
+    uint64_t _mutationCount = 0;


I can fathom a textbuffer having more than 900 quin-zill-trillion mutations. Is this a count, or an opaque number? If it's an opaque number, which is not orderable, wraparound is implicitly OK.

If somebody is going to use > on it, we're in trouble.

DHowett · 2023-08-21T18:08:34Z

src/cascadia/TerminalControl/ControlCore.cpp

-        if (text.size() == 0)
+        auto lock = _terminal->LockForWriting();
+
+        if (_searcher.IsStale() || _searcherText != text || _searcherGoForward != goForward || _searcherCaseSensitive != caseSensitive)


hmm. we need some way to tell the searcher that it is stale when we switch into the alt buffer.

The alt buffer will not have the same mutation count.

There's a one in 900 quin-jillion-millio nchance that it will, but... fair.

Yeah, I'll take that chance! 😄 I mean the worst-case scenario is that the search finds results that are stale.

asklar · 2023-08-21T20:17:12Z

Oh wait, I think you meant a different problem, right? I think that's something like Terminal::GetHyperlinkAtBufferPosition, because that one searches only a single line?

@lhecker yes this sounds about right - will this PR address hyperlinks being broken up across lines (and launching to the wrong url)? if not, is there a bug tracking that already?

DHowett

I love it. The mutable row stuff is great, and Search is completely understandable by a single human now. Thank you.

I need @carlos-zamora to sign off on the UIA changes -- my biggest concern is searching backwards off a previous search (?) because of accidental off-by-one errors. I am not sure how that API is supposed to work.

Blocking until Carlos signs, but otherwise I am a ✔️

DHowett · 2023-08-22T18:18:14Z

src/buffer/out/Row.cpp

+    const til::CoordType columnCount = _columnCount;
+    const til::CoordType scale = _lineRendition != LineRendition::SingleWidth;
+    const til::CoordType padding = _doubleBytePadded;
+    return (columnCount - padding) >> scale;


DHowett · 2023-08-22T18:20:24Z

src/buffer/out/UTextAdapter.cpp

+//
+// @param ut the UText to get the length of.
+// @return the length, in the native units of the original text string.
+static int64_t U_CALLCONV utextNativeLength(UText* ut) noexcept


should these functions return a failure if the underlying buffer was mutated since the utext adapter was created? Or is that not related to the purpose of these guys?

I feel like that this would provide a false sense of security that people might end up relying on. Unlocking the console at all during a search is likely a bug or at least a bad idea because it's pretty error prone (not all state relevant to searching is contained in TextBuffer after all - selections for instance). I think we could make it an assertion? Not sure if it should be assert() or WI_ASSERT(). Probably the former, because searching without holding the lock is still somewhat predictable.

src/buffer/out/UTextAdapter.cpp

src/buffer/out/UTextAdapter.h

src/buffer/out/textBuffer.hpp

DHowett · 2023-08-22T18:53:07Z

src/buffer/out/UTextAdapter.cpp

+
+// An excerpt from the ICU documentation:
+//
+// Extract text from a UText into a UChar buffer.


Did you complete this one?

No. I'll keep the code but remove it from the utextFuncs struct and add a comment that the correctness of the function hasn't been verified yet. I don't think it's worth it to finish that function, because we don't need it right now. But I do think it's worth keeping because the code has already been written and is probably 99% correct (if not 100%). I initially thought I had to implement it so that uregex can work, but it seems I was wrong about that. It uses the ut->chunkContents directly (via UTEXT_NEXT32 and friends).

zadjii-msft · 2023-08-22T19:08:49Z

I need @carlos-zamora to sign off on the UIA changes

Basically same here. I think I've got like 3 files left but I didn't want to ✅ without him reading that (cause obviously no one knows how that works)

carlos-zamora · 2023-08-23T16:17:26Z

src/types/UiaTextRangeBase.cpp

-    auto searchAnchor = _start;
-    if (searchBackward)
+    const std::wstring_view queryText{ text, SysStringLen(text) };
+    const auto results = _pData->GetTextBuffer().SearchText(queryText, ignoreCase != FALSE, _start.y, _end.y + 1);


Suggested change

const auto results = _pData->GetTextBuffer().SearchText(queryText, ignoreCase != FALSE, _start.y, _end.y + 1);

const auto results = _pData->GetTextBuffer().SearchText(queryText, !!ignoreCase, _start.y, _end.y + 1);

I think we do !! generally. I'm not really attached to this one though. Idk if one is better than the other.

carlos-zamora · 2023-08-23T17:07:32Z

src/types/UiaTextRangeBase.cpp

-    auto searchAnchor = _start;
-    if (searchBackward)
+    const std::wstring_view queryText{ text, SysStringLen(text) };
+    const auto results = _pData->GetTextBuffer().SearchText(queryText, ignoreCase != FALSE, _start.y, _end.y + 1);


Previously, this code could use the Search class, because Search would stop searching once it found something. The new Search class will go through the entire buffer however.

Won't this cause a decent performance hit? _getOptimizedBufferSize() used to reduce the search space by ignoring anything past the mutable bottom. Now that we're searching through the whole buffer again, won't FindText('a', ...) be slower. And if the user calls that function repeatedly, the user may have to wait quite a bit before being able to perform the next search? (FYI I haven't verified the performance question via a11y insights yet, so this could all be false)

It seems Accessibility Insights is kinda buggy? I can't really tell how it works, but if you open the Find dialog and ever reach a "not found" situation, it'll continue to say "not found" forever.

You could also use inspect.exe. It's the predecessor to a11y insights and can be found in the Windows SDK's \bin\<version>\<platform> folder. To interact with the API, do the following:

open inspect.exe

click on the terminal (the terminal should be highlighted in the UIA tree in inspect.exe)

Click "Action" > "Text Pattern Explorer..." (a new "Text Explorer" window should appear)

Click "Range" > "Find..."

Was that intentional? Should the UIA search circle around in the buffer when it has reached either end?

Interesting! I was originally going to say that we should keep the same behavior as before, but it looks like this experience is different from MS Word's! There were two notable differences with your implementation:

circling the buffer: MS Word doesn't circle the text area whereas we do in Terminal (old and new impl). So, we probably shouldn't do that anymore.

no more results: MS Word returns something that is interpreted by inspect.exe as "ProcessFind: FindText couldn't find matching text.". But Terminal (new impl) is interpreted as "ProcessFind: MoveEndpointByRange failed. Error -2147024809".

We should align these experiences (which is a bit annoying because your PR wasn't really intended to deal with any of this). If that's easy to do, it'd be best to do that as a part of this PR or in a follow-up before the next release. If not, I think we should keep the old experience, but I know that's not really possible anymore given that the Search class is now completely different. Thoughts?

zadjii-msft

I'm giving this a ✅ on the assumption that the UIA stuff is fine. I don't feel like any of these comments are blocking. And this makes searchv2 so much nicer

src/buffer/out/Row.cpp

zadjii-msft · 2023-08-22T14:56:46Z

src/buffer/out/Row.cpp

+    auto mapper = _createCharToColumnMapper(offset);
+    return mapper.GetLeadingColumnAt(offset);


it seems weird to me that you need to pass the offset into the ctor (via _createCharToColumnMapper), then immediately just call Get*ColumnAt(offset) with the same offset. I don't think I'll block over that but it feels weird

_createCharToColumnMapper only uses the offset to guess the initial column. GetLeadingColumnAt will then actually compute the proper offset. This approach avoids GetLeadingColumnAt being O(n) most of the time.

The reason the CharToColumnMapper exists as a separate class is because I want to use it for DirectWrite to map the character positions that DirectWrite returns directly back to columns. Instead, right now, we got a really ugly and inefficient approach. You can see it if you search for bufferLineColumn.

src/cascadia/TerminalCore/Terminal.cpp

zadjii-msft · 2023-08-24T15:04:05Z

src/cascadia/TerminalCore/Terminal.cpp

@@ -1348,6 +1345,97 @@ void Terminal::_updateUrlDetection()
    }
 }

+// Interns URegularExpression instances so that they can be reused. This method is thread-safe.


Okay so I thinks I get what this does, but it might deserve a bit of a longer comment than this.

If I'm not mistaken, this basically creates a process-wide cache of unique_uregexs. That way, if someone comes asking for a particular regex, (on any thread), we can hand back the pre-compiled version of it, if we've already seen it. There's also some affordance for flushing out old cache entries, if there's more than 128.

maybe intern isn't the right verb:

This is a weird one. In the compiler/language runtime field, it pretty much means means "caching". You almost always only hear it in the context of strings (e.g. "string interning").

TIL i guess. Feel free to ignore me

I turned it into a small class so we can more easily extract it in the future. I left it in the same place though because I still feel like it's better to keep this relatively "new" code close to where we're going to add custom patterns for the first time soon.

zadjii-msft · 2023-08-24T15:06:05Z

src/cascadia/TerminalCore/Terminal.cpp

+        const auto guard = shared.lock.lock_shared();
+        if (const auto it = shared.set.find(pattern); it != shared.set.end())
+        {
+            return ICU::unique_uregex{ uregex_clone(it->second.re.get(), &status) };


does this need to get std::move'd on the way out? I always forget how exactly move semantics work

(and below too)

Returned objects are always being moved, but with much stronger guarantees than regular std::move. return will behave as if there was no function call at all. In other words, returning a std::string by value is like constructing that string in the caller directly.
std::move is basically just a T& with extra syntactic sugar. So, if you return std::move(string); it will actually be forced to call the string(string&&) constructor on the way out, because you're not returning an object value anymore - you're returning a reference which now needs to be converted to a value. That's why one should never return std::move.
The reason it can be seen anyways sometimes is because when you have code like this:

struct StringWrapper { // implicit constructor! StringWrapper(std::string&& str) {} }; StringWrapper foobar() { std::string str = "..."; // calls the implicit constructor return std::move(str); }

getFontCollection() in FontCache.h is an example for that.

src/buffer/out/UTextAdapter.cpp

lhecker · 2023-08-24T21:50:34Z

src/cascadia/TerminalCore/Terminal.cpp

+    //
+    // An alternative approach would be to not make this method thread-safe and give each
+    // Terminal instance its own cache. I'm not sure which approach would've been better.
+    ICU::unique_uregex Intern(const std::wstring_view& pattern)


FYI suppress whitespace changes. The conversion to a class barely changed anything.

This is a resurrection of #8588. That PR became painfully stale after the `ControlCore` split. Original description: > ## Summary of the Pull Request > This is a PoC for: > * Search status in SearchBox (aka number of matches + index of the current match) > * Live search (aka search upon typing) > ## Detailed Description of the Pull Request / Additional comments > * Introduced this optionally (global setting to enable it) > * The approach is following: > * Every time the filter changes, enumerate all matches > * Upon navigation just take the relevant match and select it > I cleaned it up a bit, and added support for also displaying the positions of the matches in the scrollbar (if `showMarksOnScrollbar` is also turned on). It's also been made SUBSTANTIALLY easier after #15858 was merged. Similar to before, searching while there's piles of output running isn't _perfect_. But it's pretty awful currently, so that's not the end of the world. Gifs below. * closes #8631 (which is a bullet point in #3920) * closes #6319 Co-authored-by: Don-Vito <khvitaly@gmail.com> --------- Co-authored-by: khvitaly <khvitaly@gmail.com>

…row (#16775) ## Summary of the Pull Request URL detection was broken again in #15858. When the regex matched, we calculate the column(cell) by its offset, we use forward or backward iteration of the column to find the correct column that displays the glyphs of `_chars[offset]`. https://github.com/microsoft/terminal/blob/abf5d9423a23b13e9af4c19ca35858f9aaf0a63f/src/buffer/out/Row.cpp#L95-L104 However, when calculating the `currentOffset` we forget that MSB of `_charOffsets[col]` could be `1`, or col is pointing to another glyph in preceding column. https://github.com/microsoft/terminal/blob/abf5d9423a23b13e9af4c19ca35858f9aaf0a63f/src/buffer/out/Row.hpp#L223-L226

Use ICU for text search

462b8d2

lhecker added Product-Conhost For issues in the Console codebase Product-Terminal The new Windows Terminal. Area-i18n Internationalization issues (e.g.: non-US input handling doesn't work) labels Aug 21, 2023

Improve documentation

a346a2a

This comment has been minimized.

Sign in to view

lhecker added 2 commits August 21, 2023 14:59

Spelling fix

dba1629

Fix a bug and x86 woes

8146ecd

lhecker commented Aug 21, 2023

View reviewed changes

lhecker added 2 commits August 21, 2023 15:32

More x86 woes, Improve UiaTextRangeBase

5ca26aa

Fix unit tests

e0fb764

This was referenced Aug 21, 2023

Allow regex search #9228

Closed

Epic: Search v2 #3920

Open

zadjii-msft reviewed Aug 21, 2023

View reviewed changes

lhecker added 2 commits August 21, 2023 19:21

Implement backwards text iteration

ecc735d

Address feedback

56c7911

DHowett requested changes Aug 21, 2023

View reviewed changes

microsoft-github-policy-service bot added the Needs-Author-Feedback The original author of the issue/PR needs to come back and respond to something label Aug 21, 2023

DHowett reviewed Aug 21, 2023

View reviewed changes

microsoft-github-policy-service bot removed the Needs-Author-Feedback The original author of the issue/PR needs to come back and respond to something label Aug 21, 2023

Address feedback

c731eae

This comment has been minimized.

Sign in to view

DHowett requested changes Aug 22, 2023

View reviewed changes

microsoft-github-policy-service bot added the Needs-Author-Feedback The original author of the issue/PR needs to come back and respond to something label Aug 22, 2023

zadjii-msft assigned carlos-zamora Aug 22, 2023

zadjii-msft mentioned this pull request Aug 23, 2023

Refine locking around pattern tree and in TermControl #9617

Closed

carlos-zamora reviewed Aug 23, 2023

View reviewed changes

zadjii-msft mentioned this pull request Aug 23, 2023

Show number of search results & positions of hits in scrollbar #14045

Merged

zadjii-msft approved these changes Aug 24, 2023

View reviewed changes

lhecker added 2 commits August 24, 2023 19:26

Cache search parameters in Search, Improve UIA search

bf419ca

Revert unnecessary changes, Fix tests

801fc76

DHowett approved these changes Aug 24, 2023

View reviewed changes

lhecker added 2 commits August 24, 2023 23:44

Turn internURegularExpression into URegularExpressionInterner

078c7be

Fix AuditMode failures

e7bc823

lhecker added AutoMerge Marked for automatic merge by the bot when requirements are met and removed Needs-Attention The core contributors need to come back around and look at this ASAP. labels Aug 24, 2023

microsoft-github-policy-service bot enabled auto-merge (squash) August 24, 2023 21:49

This comment has been minimized.

Sign in to view

would've is not a word

2b88b6f

lhecker commented Aug 24, 2023

View reviewed changes

DHowett approved these changes Aug 24, 2023

View reviewed changes

Fix MOAR AuditMode failures

f9935f7

DHowett approved these changes Aug 24, 2023

View reviewed changes

SearchTests deserve love too, Leonard

1117450

DHowett approved these changes Aug 24, 2023

View reviewed changes

microsoft-github-policy-service bot merged commit cd80f3c into main Aug 24, 2023

microsoft-github-policy-service bot deleted the dev/lhecker/icu-utext branch August 24, 2023 22:56

This was referenced Aug 28, 2023

[1.19 selfhost] URL autodetection is broken #15891

Closed

Feature Request: Fuzzy Search #5314

Open

zadjii-msft linked an issue Sep 7, 2023 that may be closed by this pull request

Search is often broken while console process is running #1215

Open

comzyh mentioned this pull request Feb 27, 2024

Fix the hyperlink detection when there are leading wide glyph in the row #16775

Merged

lhecker mentioned this pull request Mar 24, 2024

Implement grapheme clusters #16916

Merged

carlos-zamora mentioned this pull request Feb 19, 2025

Allow newTabMenu's matchProfile to work with regex #18553

Open

		_results = textBuffer.SearchText(str, caseInsensitive);
		_mutationCount = textBuffer.GetMutationCount();

	const auto results = _pData->GetTextBuffer().SearchText(queryText, ignoreCase != FALSE, _start.y, _end.y + 1);
	const auto results = _pData->GetTextBuffer().SearchText(queryText, !!ignoreCase, _start.y, _end.y + 1);

		auto mapper = _createCharToColumnMapper(offset);
		return mapper.GetLeadingColumnAt(offset);

Use ICU for text search #15858

Use ICU for text search #15858

Conversation

lhecker commented Aug 21, 2023

Validation Steps Performed

This comment has been minimized.

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

lhecker Aug 21, 2023 • 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

zadjii-msft commented Aug 21, 2023

lhecker commented Aug 21, 2023

lhecker commented Aug 21, 2023

zadjii-msft left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

lhecker Aug 21, 2023 • 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

lhecker Aug 21, 2023 • 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

DHowett 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

asklar commented Aug 21, 2023 • edited Loading

This comment has been minimized.

DHowett 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

lhecker Aug 22, 2023 • edited Loading

Choose a reason for hiding this comment

zadjii-msft commented Aug 22, 2023

Choose a reason for hiding this comment

Choose a reason for hiding this comment

zadjii-msft 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

lhecker Aug 24, 2023 • edited Loading

Choose a reason for hiding this comment

This comment has been minimized.

Choose a reason for hiding this comment

lhecker Aug 21, 2023 •

edited

Loading

lhecker Aug 21, 2023 •

edited

Loading

lhecker Aug 21, 2023 •

edited

Loading

asklar commented Aug 21, 2023 •

edited

Loading

lhecker Aug 22, 2023 •

edited

Loading

lhecker Aug 24, 2023 •

edited

Loading