By Modules

In this mode we access the individual GMT modules directly by their name, and options are set using keyword arguments. The general syntax is (where the brackets mean optional parameters):

[output objects] = modulename([cmd::String="",] [argi=[],] opt1=val1, opt2=val2, kwargs...);

where modulename is the program name (e.g. coast), cmd is used to transmit a file name for modules that will read data from files and argi is one or, and for certain modules, more data arrays or GMT.jl data types. opti named arguments common to many modules used for example to set the output format. Finally kwargs are keyword parameters used to set the individual module options. But contrary to the Monolithic usage, the one letter GMT option syntax may be replaced by more verbose aliases. To make it clear let us look at couple of examples.

coast(region=:global, proj=(name=:laea, center=[300,30]), figsize=6, axes="g", land="navy")

This command creates a map in PotScript file called GMTjl_tmp.ps and save it in your system's tmp directory. For comparison, the same command could have been written, using the classical one letter option syntax, as:

coast(R="g", J="A300/30/6c", B="g", D="c", G="navy")

So, each module defines a set of aliases to the one letter options that are reported in each module man page.

Before diving more in the way options may be transmitted into the module, we have to understand what happens with the output image file. By not directly specifying any format we are using the default output image format which is PNG (actually, except for grdimage -A, the only format that GMT can write PostScript). But we can select other formats by using the fmt keyword, for example fmt="jpg" or fmt=:pdf. In such cases, the ghostscript program (you need to have it installed) will take care of converting the ps file into the selected format.

When runing from Jupyter notebooks one does not need to worry about the image format. In fact the only allowed is png but that is taken care automatically, meaning that any fmt=xxx will be ignored.

Note that we used either strings ("") or symbols (:) to represent the format. Here the rule is we can use symbols for any string argument that can be safely written as a symbol. Example, this is valid =:abc, but this is not =:+a (apparently parser will try to add to a). The use of symbols may be preferred for a question of laziness (less typing).

The above example, however, does not use any input data (coast knows how to find its own data). One way of providing it to modules that work on them is to send in a file name with the data to operate on. This example

grdimage("@tut_relief.nc", shade="+ne0.8+a100", proj=:merc, axes=:a, show=true)

reads a the netCDF grid tut_relief.nc and displays it as a Mercator projected image. The '@' prefix is used by GMT to know that the grid file should be downloaded from a server and cached locally. This example introduces also the show=true keyword. It means that we want to see right way the image that has just been created. While it might seem obvious that one wants to see the result, the result might not be ready with only one GMT module call. And that's why the GMT philosophy uses a layer cake model to construct potentially highly complex figures. Next example illustrates a slightly more evolved example

topo = makecpt(color=:rainbow, range=(1000,5000,500), Z=[]);
grdimage("@tut_relief.nc", shade="+ne0.8+a100", proj=:merc, axes=:a, color=topo)
colorbar!(position="jTC+w5i/0.25i+h+o0/-1i", region="@tut_relief.nc", color=topo,
          axes="y+lm", fmt=:jpg, show=true)

Here we use the makecpt command to compute a colormap object and used it as the value of the color keyword of both grdimage and colorbar modules. The final image is made up of two layers, the first one is the part created by grdimage, which is complemented by the color scale plot performed by colorbar. But since this was an appending operation we HAD to use the ! form. This form tells GMT to append to a previous initiated image. The image layer cake is finalized by the show=true keyword. If our example had more layers, we would have used the same rule: second and on layers use the ! construct and the last is signaled by show=true.

By defaultn the image files are written into tmp system directory under the name GMTjl_tmp.ps (remember PostScript is the default format) and GMTjl_tmp.xxx when user specifies a different format with the fmt keyword. It's one of this files that shows up when show=true is used. But we may want to save the image file permanently under a different name and location. For that use the keyword savefig=name, where name is relative or full file name.

The examples above show also that we didn't completely get rid of the compact GMT syntax. For example the shade="+ne0.8+a100" in grdimage means that we are computing the shade using a normalized a cumulative Laplace distribution and setting the Sun direction from the 100 azimuth direction. For as much we would like to simplify that, it's just not possible for the time being. To access the (very) high degree of control that GMT provides one need to use its full syntax. As such, readers are redirected to the main GMT documentation to learn about the fine details of those options.

Setting line and symbol attributes has received, however, a set of aliases. So, instead of declaring the pen line attributes like -W0.5,blue,–, one can use the aliases lw=0.5, lc="blue", ls="–". An example would be:

plot(collect(1:10),rand(10), lw=0.5, lc=:blue, ls="--", marker=:circle,
     markeredgecolor=0, size=0.2, markerfacecolor=:red, title="Bla Bla",
     x_label=:Spoons, y_label=:Forks, show=true)

This example introduces also keywords to plot symbols and set their attributes. Also shown are the parameters used to set the image's title and labels.

But setting pen attributes like illustrated above may be complicated if one has more that one set of graphical objects (lines and polygons) that need to receive different settings. A good example of this is again provided by a coast command. Imagine that we want to plot coast lines as well as country borders with different line colors and thickness. Here we cannot simple state lw=1 because the program wouldn't know which of the shore line or borders this attribute applies to. The solution for this is to use tuples as values of corresponding keyword options.

coast(limits=[-10 0 35 45], proj=:merc, shore=(0.5,"red"), axes=:a,
      show=true, borders=(1,(1,"green")))

Here we used tuples to set the pen attributes, where the tuple may have 1 to 3 elements in the form (width[c|i|p]], [color], [style[c|i|p|]). The borders=(1,(1,"green")) option is actually a tuple-in-a-tuple because here we need also to specify the political boundary level to plot (1 = National Boundaries).

Specifying the pen attributes

So, in summary, a pen attribute may be set in three different ways:

  1. With a text string that follows the width, color, style specs as explained in Specifying pen attributes

  1. By using the lw or linewidth keyword where its value is either a number, meaning the line thickness in points, or a string like the width above; the color is set with the lc or linecolor and the value is either a number between [0 255] (meaning a gray shade) or a color name (for example "red"); and a ls or linestyle with the value specified as a string (example: "- -" plot a dashed line).

  1. A tuple with one to three elements: ([width], [color], [style]) where each of the elements follow the same syntax as explained in the case (2) above.

Specifying the axes

The axes are controlled by the B or frame or axes keywords. The easiest form it can have is the axes=:a, which means do an automatic annotation of the 4 map boundaries – left, bottom, right and top – axes. To annotate only the left and bottom boundaries, one would do axes="a WSne" (note the space between a and WSne). For a higher level of control the user must really consult the original -B documentation.

Other than setting titles and labels with a axes string we can also do it by using the keywords title, x_label and y_label.

The figure limits is set with the R, region or limits keywords. Again, the full docs for this option are explained in -R documentation. But other than the string version, the numeric form region=[xmin xmax ymin ymax] is also permitted. And when dealing with grids, even the region=mygrid.grd is a valid operation. Where mygrid.grd is a GMTgrid type. The plot() function allows a no limits setting, in which case it will default to the data's bounding box.

Axes (and other) configuration

There are almost 150 parameters which can be adjusted individually to modify the appearance of plots or affect the manipulation of data. When a program is run, it initializes all parameters to the GMTdefaults (see more at GMT defaults). At times it may be desirable to temporarily override some of those defaults. We can do that easily by using any of the keywords conf, par or params, which are recognized by all modules. Its usage follows closely the syntax described at gmt.conf but using Named Tuples. The parameter names are always given in UPPER CASE. The parameter values are case-insensitive unless otherwise noted and can be given as strings or numeric. Provide as many parameters as you want in the named tuple. Example

basemap(...., conf=(MAP_TICK_LENGTH_PRIMARY=0.25, FORMAT_GEO_MAP="ddd:mm:ssF"))

Specifying the figure size

Figure sizes are automatically set to 12x8 cm for basic case of Cartesian xy plots done with the plot() function but otherwise in general they need to be user specified using the J or proj or projection keywords. See the full doc at -J documentation. For Cartesian plots one can also use the figsize=width or figsize=[width height] keyword, where the dimensions are in centimeters. The array form allows also set height or width to 0 to have it recomputed based on the implied scale of the other axis. Use negative sizes to reverse the direction of an axis (e.g., to have y be positive down). If neither of these forms is used, the figure width defaults to 14 cm.

The output format

It was referred above that the fmt determines the output format and that the default is PostScript. Actually, the default format is chosen by the contents of the global FMT variable set at the top of the GMT.jl file. Eventually this will evolve to using an environment variable but for the moment users will have to edit that file to set a different default format.

An interesting alternative is to set FMT="", that is to not specify any image format. This will result in NOT saving any file on disk but to keep the PS figure internally stored in the program's memory. In other words, the figure is built and kept in memory only. This allows converting to another format directly without the use of an intermediary disk file. The conversion is performed by the psconvert GMT module that would be used like this (to convert to PDF):

psconvert(in_memory=true, adjust=true, format=:f, out_name="myfig.pdf")

The issue with this solution, that could be implemented internally without user intervention, is that it currently only works on Windows.

Another alternative to a file format is the option to create RGB images with psconvert and return it to Julia as a Image type type.

I = psconvert(in_memory=true, adjust=true)

but again, so far on Windows only. A cool thing to develop would be the possibility to display this I image with the Images.jl package.

Saving data to disk

As referred in the Monolithic section, we have two programs to do read and writing. Their module names are gmtread and gmtwrite. These modules allow to import and export any of the GMT data types to and from external files. For instance, to save the grid G stored into a GMTgrid type into the file relief.nc we run

gmtwrite("relief.nc", G)

Here there is no need to inform about the type of data that we are dealing with because that can be inferred from the type of the numeric argument. There are cases, however, where we may want to save the result of a computation directly on disk instead of assigning it to a Julia variable and latter save it with gmtwrite. For computations that deal with grids that is easy. Just provide ask for an output name using the outgrid keyword, like

grdcut(G, limits=[3 9 2 8], outgrid="lixo.grd");

but for table data the GMT programs normally output their results to stdout so if we want to save data directly to disk (as would do the corresponding GMT shell command) we use the write or |> keywords. We can also use this mechanism to append to an existing file, but then we use the append keyword. Use together with the bo option to save as a binary file. The following converts the grid G to x,y,z triplets and save the result in an ASCII disk file.

grd2xyz(G, write="lixo.xyz")

How inputs are transmitted to modules

Different modules take different number of inputs (for example grdblend accepts a variable number of grids) and some modules accept primary input and optionally a secondary input (for example the weights option in grdtrend). The primary input(s) can be sent as text strings with the names of files to be read or as Julia variables holding the appropriate data type, and that as the first argument to the module call. Alternatively, the numeric input can be sent via the data keyword whose value can be a tuple when the expected input is composed by more than one variable. The same applies when an option is expected to receive more than one arguments (for example the three R,G,B in grdview). Examples:

grdimage(G, intens=I, J=:merc, color=C, B="1 WSne", X=:c, Y=0.5, show=1)

grdimage(data=G, intens=I, J=:merc, color=C, B="1 WSne", X=:c, Y=0.5, show=1)

grdview(G, intens=:+, J=:merc, zsize="2i", view=(145,35), G=(Gr,Gg,Gb), Q=:i, show=1)