Repository navigation
Feature/sparse chunk cache - #211
Merged
Merged
Conversation
…onditional cursor) Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…or, simplify guard) Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- suppress the click synthesized after a resize drag so finishing a resize (even past the min-width point) never triggers a column sort - in resizeColumns, set column.width to the natural header width when there are no data cells, so a resize on an empty grid does not snap the column width Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Save bodyViewport.scrollLeft before clearDOM wipes the DOM and restore it after the body is rendered, so sorting/paging no longer jumps the grid to 0. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Column resizing and sorting in EasyGrid
Two width-stability fixes for EasyGrid: - Always render the sort-indicator box (fixed 12px, empty when unsorted) on sortable headers, so the arrow appearing/disappearing no longer changes the header's natural width used by the auto-fit lower bound. - Make AutoResizeColumns.Once survive refetches: the grid keeps a map of auto-computed widths per column id and carries them over to the fresh DataColumn instances each server response brings, so columns are measured once on first render instead of being re-fit to every new page of data. Always mode is unchanged. Covered by 4 new latte tests in easy_grid_interactions.test.ts. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The header label is a flex item, so without min-width: 0 it refuses to shrink below its longest word and the text bleeds into the neighboring header cell when a column is resized down to its minimum. Let it shrink and clip the overflow with an ellipsis (data cells already do the same). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The mobile UI (full-page menus instead of drop-downs, no drag and drop, no hover-driven interactions) was chosen by the window size alone, so any window narrower than 840px got it - including a narrow desktop window, a browser side pane or a half-screen window, where it only gets in the way. The screen size is still required, but now the device must also have no hover and a coarse primary pointer. Browsers which don't understand those media features report the query as "not all" and fall back to the size check, and setIsMobileMode() keeps overriding everything. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
…inned widths), the allowColumnResize option and the column resize event. [new]: EasyGrid: clickable sortable header with a sort indicator, the sortable option and the columnSort event, plus per-column sort direction / width state and helpers. [upd]: The mobile mode is turned on for touch devices only - besides the screen size the device must have no hover and a coarse pointer, so a narrow desktop window or a browser side pane keeps the normal UI (the size check remains as a fallback, setIsMobileMode() still overrides everything). [fix]: The horizontal scroll position of the grid is preserved across re-renders (sorting, paging). [fix]: The header label is clipped with an ellipsis at narrow column widths instead of bleeding into the neighboring header cell. [fix]: An exception was thrown on the conversion to text for numeric fields in FullTextSearchQuery (EF Core 5+ uses the instance ToString(), on EF Core 2/3 such properties are skipped). [fix]: Numeric fields are included into the full text filter only when the search text is a number.
The rows used to be cached as one contiguous run starting at the very first one, so reading a distant page pulled in everything before it - jumping to page 200 meant loading pages 1 to 200. The rows are now held as a set of disjoint segments (SparseRowCache), and a read only asks the loader for the parts that are really missing. The requested windows are aligned to chunkSize, so reading any row of the same chunk asks for the same window and it is loaded once. The windows that are not adjacent to the first row are dropped on an LRU basis once there are more than maxCachedWindows (8 by default). getCachedRows() and getCachedCount() keep their meaning - the rows loaded in one uninterrupted run from the very first one. That run has to stay gap-free and in order, as the aggregate calculators, the pivot table and the charts walk it from the beginning. The rows of a detached window are deliberately not exposed there. Also: - the chunk requests for the same window are shared instead of being sent twice, and a chunk requested before the rows were dropped is no longer merged into the table it no longer belongs to; - "the total is known" is told apart from "the total has been asked for". They were one flag, so a read made while the first one was still on its way saw the total as 0 and returned no rows at all; - ChunkInfo is exported as an alias of DataChunkDescriptor - the tests already referred to the chunk parameters under that name. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
This pull request introduces a new sparse row caching mechanism to the
EasyDataTableclass in the EasyData.JS core package. The main goal is to improve memory management and performance when handling large datasets by caching only needed windows of rows and evicting older ones as necessary. The changes also add configurability for the cache size and refactor row loading logic to support these improvements.Key changes include:
Sparse Row Cache Integration
SparseRowCacheclass inEasyDataTable, allowing for non-contiguous, windowed caching of rows. This enables efficient memory use and supports loading and evicting windows of rows as needed. (easydata.js/packs/core/src/data/easy_data_table.ts, [1] [2] [3]maxCachedWindowsoption. (easydata.js/packs/core/src/data/easy_data_table.ts, [1] [2]Row Loading and Window Management
planLoadWindowsandloadWindowmethods to manage this process. (easydata.js/packs/core/src/data/easy_data_table.ts, easydata.js/packs/core/src/data/easy_data_table.tsL146-R290)easydata.js/packs/core/src/data/easy_data_table.ts, [1] [2]API and Type Improvements
maxCachedWindowsoption toEasyDataTableOptionsand provided documentation for its use. (easydata.js/packs/core/src/data/easy_data_table.ts, easydata.js/packs/core/src/data/easy_data_table.tsR15-R37)ChunkInfotype alias forDataChunkDescriptorfor backward compatibility. (easydata.js/packs/core/src/data/data_loader.ts, easydata.js/packs/core/src/data/data_loader.tsR13-R15)Miscellaneous
1.6.0-rc01to reflect these significant changes. (easydata.js/packs/core/package.json, easydata.js/packs/core/package.jsonL3-R3)These changes collectively make row access more efficient and scalable, especially for large datasets, and provide more control over memory usage.