Skip to contents

Changelog

Source: NEWS.md

urbioconnect (development version)

  • The shiny app can be shipped as a desktop installer for macOS and Windows, built by shinyelectron from a pushed version tag. The installer carries its own R and its own Quarto, so someone double-clicking it needs neither installed. Nothing in the package changes: the installer is assembled entirely in CI. (#147)

  • run_connectivity_app() now checks for diffviewer before launching. The app attaches it for the Scenario tab’s compare view, so without it the check passed and the app then failed to start. The check and the tests’ skip list are now one vector, which is how they came to disagree.

  • The shiny app offers two example datasets rather than one. The Blue Tongue Lizard landscape is about 200x200 cells and analyses in under a second; the Superb Fairy Wren one is about 1500x1400 and takes tens of seconds, so which you are asking for is now a choice rather than a tick box. Each dataset brings its own scenarios, and the scenario dropdown offers only the ones that go with the landscape in use.

  • New example_barrier_scenario() gives the lizard barrier with a new road cut through it, so the small example dataset has a scenario of its own. It is derived from example_barrier() rather than shipped as a second file.

  • The shiny app takes an optional scenario layer alongside the habitat and the barrier: one of the supplied scenarios, or your own upload of a changed habitat or barrier. Run Analysis then compares it to the baseline, and the Results tab gains a Scenario tab holding the comparison as one row per metric, in the shape the analysis produces it, and as maps of the two landscapes. The maps get a Compare view built on diffviewer, the widget testthat::snapshot_review() uses, so the two landscapes can be read as a pixel difference, a toggle or a slider, at one tab per interpatch distance. The comparison travels into the report and the download. (#35)

  • The shiny app gains a Reports card, with buttons for the HTML report, the PDF report, and both as a zip. The two report buttons existed but had no handlers, so they did nothing. Without the Quarto command line tool the three are disabled and the card says where to get it; every other download still works. (#61, #62)

  • Warnings about a distance that the raster resolution can’t represent now talk in interpatch distances, matching the argument you supplied.

  • Functions taking a dir to fill create it if it doesn’t exist; functions taking a path to one file error if that file’s directory doesn’t, as ggplot2::ggsave() and readr::write_csv() do. generate_connectivity_report() therefore no longer creates its output directory, which the old output_dir argument did.

  • New connectivity_display() is the one place that decides what a metric is called and how precisely to show it, with round_by() and format_by() to apply that. The report and the shiny app both use it, so they no longer print different numbers for the same analysis: effective_mesh_ha used to show three decimal places on screen and two in the report. A comparison displays wide by default, one row per metric and a column per measure, which is the shape a reader wants; wide = FALSE keeps the object’s own shape. display_ids() says which columns identify a row, so a caller can pivot or drop them without knowing the label text.

  • New connectivity_file_stem() gives the species-and-date name every download takes, so a report, a zip and the folder inside that zip agree.

  • New onto_grid() puts a layer on a given grid, rasterising a vector one and resampling a raster one, and is the one place that decides what an empty cell means. prepare_rasters() uses it for both of its layers and so now accepts a SpatRaster as well as an sf layer: the shiny app offers GeoTIFF uploads, which used to fail inside terra::rasterize().

  • connectivity_report_data() gains scenario and scenario_kind, so one call runs a baseline and a scenario and returns both landscapes with the comparison between them. It replaces the comparison argument, which took a compare_connectivity() built elsewhere and so ran the baseline pipeline a second time: on the example wren data at three distances this is 43 seconds rather than 77, for the same numbers. The scenario is put on the baseline’s grid on the way in, so it need not arrive on one, and only one layer may change at a time, which scenario_kind now makes structural rather than an error. The report gains a scenario section and the download a scenario-comparison.csv; without a scenario, neither appears. (#35)

  • New scenario_layer() returns the layer a scenario changed, so a caller doesn’t branch on scenario_kind itself.

  • New generate_connectivity_report() renders one document holding the maps, tables and summary for an analysis. The format comes from the file extension, as it does for ggplot2::ggsave(): generate_connectivity_report(x, "report.pdf") writes a PDF, "report.html" an HTML file, and any other single-file format Quarto can write, such as .docx, also works. A .pdf goes through Typst, so no LaTeX is needed, and the figures are drawn from the data rather than embedded as images. Needs the Quarto command line tool. (#54, #61)

  • New gg_layer() draws one layer on its own, the habitat or the barrier, rather than combined as gg_barrier_habitat_interpatch_dist() draws them. It takes the colours from the palette by kind, so a barrier is drawn as it appears on the combined map: white, against the interpatch green, because white on white would be nothing at all. The shiny app’s Scenario tab uses it to show the layers an analysis was built from.

  • gg_barrier_habitat_interpatch_dist() and plot_barrier_habitat_interpatch_dist() now default col_barrier, col_interpatch_dist and col_habitat to the package palette, so the app, the downloads and the report can’t drift apart. (#54)

  • habitat_connectivity(), habitat_connectivity_full(), sf_habitat_connectivity(), habitat_connectivity_comparison(), habitat_connectivity_scenarios() and connectivity_report_data() now take interpatch_distance only. The buffer radius is half the interpatch distance, so offering both was two ways to say the same thing; habitat_buffer() and sf_habitat_buffer() still take buffer_radius, since that is the operation they perform.

  • habitat_connectivity_comparison() now errors when the scenario and the baseline are not on the same grid. Comparing layers of a different extent, resolution or CRS used to succeed, and reported the change of place as a change in connectivity. (#35)

  • plot_barrier_habitat_interpatch_dist() now saves at the shared figure size rather than whatever size the last graphics device happened to be, and slugifies the species in the file name, so a species with a space no longer produces a file name with one. (#54)

  • New quarto_available() says whether the Quarto command line tool is installed, which is what the reports need and nothing else does. (#61)

  • New urbio_figure_size() gives the figure width, height and resolution the downloadable PNGs and the report share. (#54)

  • write_connectivity_assets() and zip_connectivity_assets() gain a reports argument, and include the HTML and PDF reports by default. It defaults to quarto_available(): without Quarto the maps, tables and GIS layers are still written and a message says the reports were skipped, so a missing Quarto costs the reports rather than the whole download. (#61, #62)

  • New write_connectivity_report() and render_connectivity_report() split report rendering into its two steps, so the Quarto source can be kept and changed rather than thrown away. write_connectivity_report() writes the document and the analysis beside it, and the pair renders on its own with the Render button or quarto::quarto_render(), giving the same figures the package does. generate_connectivity_report() is now the two of them in one call. (#61)

  • New write_report_data() and read_report_data() save a connectivity_report_data() to a file and read it back, packing the rasters so they survive the trip. This is how an analysis reaches the separate R session the report renders in. (#54)

  • New zip_connectivity_reports() archives just the two reports, for someone who wants the write-up without the GIS layers. (#62)

  • The shiny app gains a “Download everything” button, returning one archive of every map, table and GIS layer, and its analysis now builds a single connectivity_report_data() rather than assembling the pieces itself. (#153)

  • New connectivity_report_data() holds everything one analysis produces: the summary for every interpatch distance, plus the buffered habitat and patch-ID raster each distance produced. (#153)

  • New write_connectivity_assets() writes the downloadable assets - maps, tables, and patch polygons as GeoPackage and shapefile - in folders by interpatch distance, with a README describing each file and the run that produced it. (#153)

  • New zip_connectivity_assets() archives those assets under a folder named for the species and date. (#153)

  • New urbio_colours() and format_resolution(): the map palette and the raster resolution formatting, so the app and the downloads can’t drift apart. (#153)

  • urbioconnect now requires terra >= 1.8-70. Earlier versions of terra::identical() ignore NA cells, so a scenario that only moved NA cells would compare as unchanged. (#140)

  • Use GPL (>= 3) License.

  • drop terra_ prefix and move rast_ functions into scratch where we test the LOO method. * Add sf_ prefix to vector based approaches.

  • Add datasets and dataset loading function

  • Add legend to habitat buffer barrier plot - #66

  • Resolve internal issue where raster might not be exactly aligned, add internal function align_to() in drop_habitat_under_barrier(), fragment_habitat(), and assign_patches_to_fragments().

  • update effective_mesh_size() and connectivity_probability() to go from area_squared –> area_baseline. #128. This will help facilitate #124.

  • habitat_connectivity(), habitat_connectivity_full(), and sf_habitat_connectivity() now take either interpatch_distance or buffer_radius (supply exactly one). The lower-level habitat_buffer() and sf_habitat_buffer() take buffer_radius directly. (#131)

  • habitat_buffer() now warns when the buffer radius is too fine for the raster resolution (smaller than one cell, or not a clean multiple of it) and returns the habitat unchanged instead of erroring. (#131)

  • New vignette vignette("interpatch-distance-and-resolution") on how interpatch distance, buffer radius, and raster resolution interact.

  • Fix the shiny app’s interpatch-distance input, which was read under the wrong id. (#131)

  • Add “patch_size” class to habitat_connectivity() function, to pave the way for attaching useful metadata to these data. (#133).

  • Add patch_size S3 class:

    • add pc_* accessor functions to get: interpatch_distance, patches, res, species.
    • Extend patch_size onto tibble
    • Add various checking functions to ensure you can compare the same species, and metrics together.
    • change area parameter for summarise/compare_connectivity to be connectivity
    • remove use of arguments, target_resolution and aggregation_factor from many functions as it is only really relevant to the spatial processing, and we really only care about the resolution at the end of the day

  • habitat_connectivity() now returns a one-row connectivity-class landscape summary (n_patches, effective_mesh_ha, prob_connectedness, patch_area_mean, patch_area_total_ha, …) instead of the raw per-patch table. The per-patch areas travel with it in a patch_size list-column, retrievable with the new patch_sizes() accessor. sf_habitat_connectivity() still returns the per-patch table directly for now. (#141)

  • Rename the per-patch class and constructor patch_size() -> patch_size_tbl() (and internal new_patch_size() -> new_patch_size_tbl()), freeing up the patch_sizes name for the new accessor above. (#138)

  • New habitat_connectivity_comparison() compares a scenario against a baseline, for one or more interpatch distances (or buffer radii). Only one of habitat or barrier may differ from the baseline, so any change can be attributed to that layer. It errors if both differ, and warns if neither does. (#140)

  • New example data example_wren_habitat_scenario(), a habitat scenario to pair with example_wren_habitat(). (#140)

  • summarise_connectivity() and habitat_connectivity() now return metrics at full precision. prob_connectedness was rounded to 6 decimal places, which was coarse enough to hide the small changes a scenario produces. (#140)

  • Argument checks use call and arg to name the function and argument that failed, and that they are called in.

  • compare_connectivity() gains a scenario_name argument: an optional label, e.g. “Scenario A”, that appears as the first column on every row. It is NA when not supplied, so labelled and unlabelled comparisons stack. (#35)

  • New compare_scenarios() compares several scenarios against one baseline, taking a named list of connectivity objects where the names become the labels, and returning four rows per scenario. (#35)

  • habitat_connectivity_comparison() gains scenario_name, passed through to compare_connectivity(). (#35)

  • New habitat_connectivity_scenarios() compares several scenarios against one baseline starting from layers, taking named lists of habitat and barrier scenario layers. A scenario changes exactly one layer, and the baseline is computed once per distance and shared by every scenario. (#35)

  • summarise_connectivity()’s default method now stores a patch_size_tbl in patch_size, the same as its patch_size_tbl method, so compare_connectivity() works on a connectivity object built from a plain vector of areas. (#35)

Breaking changes

  • buffer_radius is gone from every function above habitat_buffer(). Code passing buffer_radius = r should pass interpatch_distance = 2 * r.

  • interpatch_distance is now the full edge-to-edge distance below which two patches count as connected. It is halved internally to the buffer radius, so connectivity results differ from previous versions; reproduce old output by passing buffer_radius = the old value. (#131)

  • habitat_connectivity() return type changed from a per-patch patch_size_tbl to a one-row connectivity summary. Code relying on per-patch columns (e.g. habitat_connectivity(...)$area) should instead use patch_sizes(habitat_connectivity(...))[[1]]. (#141)

  • compare_connectivity() is now compare_connectivity(scenario, baseline). It takes two one-row connectivity objects, such as the output of habitat_connectivity() or summarise_connectivity(), and returns four rows: baseline, scenario, change (scenario minus baseline) and pct_change. It is no longer an S3 generic, and its patch_size_tbl and default methods are gone. (#140)

  • summarise_connectivity() no longer takes connectivity_baseline. Use compare_connectivity() to compare against a baseline instead. (#140)

urbioconnect 0.1.0

  • Make a NEWS file to monitor changes.