CURRENT PMESH ISSUES OPEN - issues that have not beeen resolved Issue: On the cluster, if storm_surge/pmesh is in the pythonpath (and not storm_surge?), pmesh.mesh fails by trying to load pmesh visualisation stuff. Pmesh shouldn't be trying to load anyhow. Issue: crash if file>filter alpha boundary if there is no alpha boundary. Issue: Add export seg outline file back into pmesh Issue: triangles_to_polySet (outlining the current set) sometimes duplicates vertices. Now that is fixed, but segments that overlap (ie one big line that gets little ones drawn on it) break triangle. Suggested action: make a dictionary of Hough transforms (parametric lines i.e. angle, x-intercept) of the segments. Check any that take up the same space. Fix overlapping lines. Issue: Red mesh selected for refinement seems to be 'behind' the green mesh after selecting all then clearing the selection. Importance: low Suggested Action: Fixed BUT now it draws the greens behind the reds. It is much less obvious now though. The only way to fix it for good is to colour in the triangles - that is the only way to guarantee you know what each triangle is. Issue: Ole thought that the regional mesh max areas were incorrect when a .tsh file was loaded. Maybe it was that the overall max area is not saved. It's hard coded in MeshGenDialog. Importance: Medium Suggested Action: Fix when .tsh goes NETCDF Issue: Catch error if .tsh file is a bad format. suggestion: quick and dirty try - except Importance: Medium Issue (new feature): Have visualisation of verts be turned on and off. This will speed things up. Importance: low Issue (new feature): Hve .tsh as a netscf fie, instead of an ascii file. Importance: Medium Issue: When a file is saved as and then saved later, sometimes the file name is not known. Is it that 'clear' is removing the name (which it should). Issue:A .tsh file loaded into pmesh after smoothing will not have the outline info, since this is removed when smoothing. If this is a problem, refactor loadASCII so it can write the outline as well. Suggested Action: Document use Issue:vertices at the bottom of the canvas can be hard to visualise, since you don't know if all vertices are on the screen. Suggested Action: have a 5% border Importance: low to not do Issue: The mesh produced by triangle isn't of high quality near a low angle formed by two user segments. Reason:triangle Tried: Tried the triangle flags L,i,V (for info). Using box.poly on nautilus. Couldn't get a better quality mesh. Issue: Get the naming right for mesh/triangulation/user/generated. How about "outline + triangulation = mesh" Importance: Medium -low Issue: S/W - too many methods with the same name on differnent classes eg editWindow. Makes things confusing. Importance: low Issue: Normalise seems broken - what is it doing to the attributes?. Suggested Action: remove! Bug: Calling 'Generate Mesh' on empty document or If there are no enclosed regions will cause trianglulation to crash pmesh. Importance: low - not do, but document Suggested Action: Catch them (with try) and do nothing notes: Fixing the 'Generate Mesh' crash is a P3 requirement. The zoom in/out is jittery The initial mesh window coordinates are in the fourth quadrant. Deleting a selected object is done with the scroll bar, not the delete key. Issue: triangle calculates incorrect attribute values, some of the time. Importance: Medium Suggested Action: Preprocess data to smoth it out. Issue: pmesh can not be automatically installed in windows and linux, triangle doesn't compile easily. Installs best in Linux. Importance: Low Suggested Action: delve into setup more so it works with a variety of compilers. Issue: two vertices can be ontop of one another Importance: Low Suggested Action: See how it goes when used. The cost of checking probably doesn't outwiegh the benifit. action taken: When loading an xya file duplicate vertices are removed. Issue: When calculating mesh, have the pointer turn into an hour glass Importance: Low **************** CLOSED - issues that have beeen resolved ************ Issue: If some triangles have tags and others don't then the data structure (a list of lists) is dificult to use in Numeric. Solution - use a list structure, instead of list of lists Closed - used a list structure, instead of list of lists Issue (Ole): I couldn't get pmesh to write mesh in the .msh format Importance: Mid Suggested action: Either 1: Equip scripted mesh generation with functionality for tagging 2: Allow pmesh to store as msh 3: Have a tsh2msh converter Closed: Couldn't be reproduced. suggested action 2 has already been implemented. More investigation required before calling an issue an issue.. Issue: After selecting all for refinement then clearing the selection segments (blue lines) can not be selected. Importance: High Suggested Action:Fixed Reason: Issue: Refined triangles don't have the correct region information. Importance: High Suggested Action: Fixed Reason: I don't think 'refine' fixes triangle attributes Issue: Pmesh won't save a refined mesh. Importance: High Reason: Refine_triangle.getNewVertex was not appending to mesh.meshVertices, nor was it assigning an index to the vertex. Issue When starting generate mesh, the cursor doesn't start in the first box. Importance: Low Action: fixed Issue: Why is triangle att's set as a list and returned as a list of lists, when loading from a .tsh file? Note: a change of this might also affect the 'triangle' glue code. Note: This also happens to triangle atts! (?) action: triangle att's set as a list of lists Issue:Implememt the following with mesh gen minimum angle, 0 - 32 nothing;32-40 warning message, may not converge; 40 - error message Importance: Low Action: done Issue: Middle magnifying glass icon does not describe what it will do. Importance: Low Low action: too unimportant to do Issue: Maybe triangle doesn't normalise meshes away from the origin, resulting in round-off error. Importance: Low Low Suggested Action: closed since this didn;t seem to be a problem. Issue:hull.exe generates files that it doesn't clean up eg called s1f8, se4, etc. These seem to turn up in a root directory Importance: Low Issue: Hull.exe (alpha shape) - gives an assertion error if points are too close together?. Importance: Low Suggested Action: when using the results from hull.exe, check if it has given any results and throw a friendly message if there are no results. Issue: When adding segments, vertices are hard to select (with the mouse). Issue: Calling autoSegment when not in the pmesh directory will not work, since it calls the hull.exe, assuming it's in the current dir Importance: medium Suggested Action: Have the current dir change to the pmesh dir , run hull, then change back action: contractor is making alpha shape in python Issue: In the shallow_water functions for pmesh, make the volume.interpolate_cons_quantitie call more automatic Issue: Need tests for pmesh functions in shallow_water.py Issue: If there is a large scale grid, how many triangles can be generated? with a normalised grid there's a limit of 22674. Using a 100 * 100 domain, a max triangle area of 0.01 produced 1,048,576 triangles. Issue: Internal boudaries get an empty string as tag (per default?) Background: The notion of defaiults values for internal boundaries was removed, but this one seems to have taken its place. The following hack will remove the tag: #for b in domain.boundary.keys(): # if domain.boundary[b] == '': # del domain.boundary[b] # #domain = Domain(domain.coordinates, domain.triangles, domain.boundary) Importance: high Issue: The default region button cannot be selected after loading a file. Reason: Unknown. Importance: Low Issue: The vertices were being sorted when loading .xya files Time to complete : 5hrs Issue: In load ascii all text reading of lines using find must ltrim beforehand Issue: pmesh -when loading/saving the dir should be the dir where the file lives. Not always the current dir. Issue: Tags for specifying boundary conditions should be arbitrary Background: Domain has the functionality for using arbitrary keys when associating boundary objects to boundary values. Triangle, which pmesh is based on, uses integers only. It would be nice to have mnemonic names for boundaries throughout. Importance: Mid Status: Unresolved Issue: Calling save the first time seemingly does nothing Importance: Low Suggested Action: Call save_as when save is clicked the first time. Bug: Calling Zoom functions on empty document causes an AttributeError Exceptions. Importance: Medium Suggested Action: Catch them (with try) and do nothing notes: the zoom crash is not reproducable. Fixing the 'Generate Mesh' crash is a P3 requirement. When doing mesh generation, no default values (min angle, max area) are shown. - Low priority for Chris Cursor shape changes when it shouldn't. Issue: Help message and status message are the same. Remove one Importance: Low Suggested Action: Remove the status info Issue: Selecting nodes require that boundary is selected precisely. Importance: High Suggested Action: Allow click in interior of square. Bug: Entering negative value for area is allowed. Importance: Medium Suggested Action: Check input and return error message Issue: Minimum angle needs a specification of what interval is allowed. It seems to pick a default value if angle is out of bounds. Action: angles <0.0, >90.0 not allowed Bug: Entering zero for area causes crash. Importance: Medium Suggested Action: Check input and return normal error message Issue: Re-generating grids goes slower and slower when repeated. Importance: Low Suggested Action: Check that old grid gets removed from memory. Action taken: cleaned up how visualised objects are handled Issue: nothing says what mode(seg or vertex) pmesh is in Importance: Low Action taken: Have the buttons stay 'down', to depict mode. Issue: Meaning of 1 x zoom is not intuitive. Related issue: Absolute scale should be reported in status bar Importance: Medium Action taken : Zoom x 1 gives a zoom corresponding to the opening screen. Vertices and segments that have a triangle on top of them are hard to select/see. Ole suggested that there be the capability to view/hide the mesh. Action taken :put segs/vertices on top NOTES The arrow and postscript bottons currently do nothing. ----------------------------------------------- NEW REQUIREMENTS Introduce 'eraser tool'. When selected left clicking any object will remove it. Allow generated grid to be cleared (or hidden). CVS INFORMATION FOR THIS FILE $Source: /var/lib/cvs/pmesh/documentation/issues.txt,v $