v1.8
Bugs fixed:
• Fixed crash: geographic locations at 0 degrees could crash, depending on your display settings.
• Fixed crash: right-clicking on an empty area of the Folders preferences tab and choosing any menu item would crash.
• Fixed crash: "Detect Data In Text" would crash if it found address or transit data in your text.
• Fix a bug where grouping by a field with Nulls would fail to group them. (Null in Strukt is essentially a singleton, not a strange non-value like in SQL. We're not constrained to equal-typed tuples, so we can just omit a value when it's missing.)
• Fix an issue with database optimization which could cause an insignificant edit to cause a (pointless but harmless) re-run.
Performance improved:
• For "Read File", use a newer API (DispatchIO, versus NSInputStream) for reading data. It's much faster than the old way.
• When an unknown file is dropped on Strukt and we need to figure out if it's a SQLite database, load less of it (only 1KB, versus 4KB). We use libmagic for this, and it checks way less than 1KB, so this should make no practical difference.
• Minor improvements to performance for sorting and grouping operations.
User interface tweaks:
• When an unknown URL scheme is entered, list all the recognized schemes in the error message.
v1.7 [Nov 16, 2020]
Output view improvements:
- Improve performance of the line chart for large datasets.
- On right-click on table -> "Open with" -> "Other...", use a sheet, instead of a floating dialog. Also, if the chosen application refuses to open the file, report that to the user.
Output view bug fixes:
- Fixed right-click on table -> "Open with" -> "Strukt". A recent bug fix inadvertently disabled this feature.
- Fixed a bug where the histogram could disappear after it finished drawing.
Operation improvements:
- Improve warning messages (especially for "Literal Value") to explain what flag formats are supported.
- Add some keyword aliases to the "Read from URL" operation, like "download" and "http", to make it easier to discover.
- In "Parse Archive" docstring, say where the data comes from ("bytes"). It was consistent with other operations that take binary data, but never actually stated.
Operation bug fixes:
- Fixed a bug which could affect the "List Files" operation. Previously, it could fail to report results if you typed a folder name which didn't end with a trailing / character.
- Fixed two separate bugs where malformed database connection URLs could crash the application.
- Fixed the "Replace Text" operation's "case-sensitive?" flag. Due to a dumb mistake, this flag was previously ignored.
- Fixed a possible bug in "Group by". I'm not sure this ever returned incorrect results, as I was not able to demonstrate or detect the bug in any actual pipelines.
Internal improvements:
- Various internal cleanups, especially related to database queries.
- Remove some old logging (mostly added in v1.7.1 for tracking down bug #39).
