* Bugfix: the entire width of all fields of edited columns were loaded, instead of a limited view.
* Bugfix: (follow-up to above) one too many rows were included in RootNodeCount after less rows were matched than expected from the data source.
* Bugfix: some memory was not released as early as it could've been when less rows were matched than expected from the data source.
Hopefully fixes issue #730.
* Indicate in CSV files which rows need to be UNHEX()ed with LOAD DATA INFILE during import. Takes care of export part of issue #709.
* Optimization: stop copying records and arrays to stack where appropriate. See also note in helpers.pas.
* Factor out code to extract a cell in it's full width into EnsureFullWidth().
* Create method DiscardData() to release memory for a row.
* Bugfix: GridTo{Xml,Csv,Html} did not export cells in their full widths; they do now. Fixes remainder of issue #685.
* Bugfix: remove superfluous field separator at end of each CSV record.
* Bugfix: CSV field enclose characters within field data must be escaped by printing the enclose character twice, as per de-facto (is it a standard when there are hundreds of applications supporting it, but *all* the major ones all come from the same company?..) CSV.
* Indicate in CSV files which cells are NULL by removing field enclosing (non-standard MySQL CSV extension).
* Indicate in XML files which cells are null and which cells are in hexadecimal format.
* Bugfix: always escape <, &, > in HTML and XML files.
Bugfix: HexToBin() will simply stop when hitting invalid characters, make sure the buffers are valid in that case (should add code to raise an exception).
* Add a few notes about where AVs seem to randomly occur when GridPost{Insert,Update,Delete} and EnsureDataLoaded invokes the query logic from outside the main thread.
Fixes issue #659 as far as the grids go, but not for export functions.
Pros:
- CSV/HTML/XML exports are now encoded as UTF-8, not UTF-16. Importing such CSV files should now work magically when using the Import CSV dialog.
- Exported files should now be significanty smaller when most content is latin1 encoded.
- Use TMemoryStreams for copy actions, TFileStreams for export actions. Should save some memory when saving to files and improves performance by factor ~100 on larger grid results.
- Fetch text from the Grid.Text, not from the underlying TGridResult.Rows array. Fixes issue #685 .
- Removes non functionial export of image files from BLOBs
- Sanitize helpers function from GUI code (mousecursor, opening result file, ...)
Todo:
- Limit export to a user defined number of rows. Important for large tables.
- Fetch entire TEXT field content, currently limited to 256 chars
Affected issues: rfe #618, rfe #464, bug #324, bug #635 (maybe more)
Breaks: Unicode grid editing (due to yet unsolved bug #620), enum pulldowns, MEMO editing
Major changes:
- editor tab at bottom completely removed, SQL log is always visible now.
- memo, blob and enum editors don't have a replacement yet. This is left for later commits.
- grid cells are colorized to signalize their datatypes
- UPDATEs, DELETEs and INSERTs are composed by own code, no longer by Zeos internal mechanisms. Should make debugging easier.
- UPDATEs + DELETEs on tables without unique key are blocked now.
- editing in the query grid is always blocked. Would need some extensive parser to allow that.
- Open the file in "read and leave" mode, don't lock it for write operations. This fixes opening a file which is opened and write locked in another application
- Handle exceptions for ReadTextfile in QueryLoad which was malfunctioning yet, lead to an AV.
- Enhance inline documentation
- OpenTextfile
- ReadTextfileChunk
- ReadTextfile
Enables us to use the unicode detection in places where we need to read a file chunkwise. (e.g. the runsqlfile unit)
TODO: Fix UTF-8 detection, misdetected as ANSI currently.
