[bug#40757] New package: VisiData

[Jack Hill <jackhill@jackhill.us>]

Ryan,

On Wed, 22 Apr 2020, Ryan Prior via Guix-patches via wrote:

> Hi Guix! This patch adds a package for VisiData.
> 
> VisiData is an interactive multitool for tabular data. It combines the clarity of a
> spreadsheet, the efficiency of the terminal, and the power of Python, into a
> lightweight utility which can handle millions of rows with ease.
> https://www.visidata.org/, GPLv3

Thanks for the patch! This looks like some cool software.

I'm not an expert reviewer, but I'd like to try to help by offering the 
following suggestions for improvement.

>diff --git a/gnu/packages/visidata.scm b/gnu/packages/visidata.scm
>new file mode 100644
>index 0000000000..d9df94dc9b
>--- /dev/null
>+++ b/gnu/packages/visidata.scm

When adding new modules, please also add the file to gnu/local.mk

>+(define-module (visidata)

The module name should be (gnu packages visidata) to match the filesystem 
path.

It might be helpful to catch problems like this to try to apply your patch 
apply your patch to a guix source checkout, and try building your package 
from there. The manual has some documentation about that, 
<https://guix.gnu.org/manual/en/html_node/Building-from-Git.html>, but I 
can answer questions as well.

>+(define-public visidata
>+  (package
>+   (name "visidata")
>+   (version "1.5.2")
>+   (source (origin
>+            (method url-fetch)
>+            (uri (string-append "https://github.com/saulpw/visidata/archive/v" version ".tar.gz"))

The GitHub archive URIs are not stable and could result in a different 
hash in the future. Often in these cases we use git-fetch, but since 
visidata is available on pypi, I would recommend

(uri (pypi-uri "visidata" version))

>+            (sha256 (base32 "0h7hq6bnc8svkcc9995kkmgcb9n5qgm85rsshzzdicmg9rg3ymhi"))))
>+   (build-system python-build-system)
>+   (arguments '(#:tests? #f))
>+   ;; Tests disabled because they are not packaged with the source tarball.
>+   ;; View test status here: https://circleci.com/gh/saulpw/visidata/tree/stable
>+   ;; Upstream suggests tests will be packaged with tarball around 2.0 release.

Nitpick: I would prefer to have the comment before the code. I think the 
line about circleci could be removed, but the other two seem good.

>+   (native-inputs
>+    `(("python-dateutil" ,python-dateutil)
>+      ("python-fonttools" ,python-fonttools)
>+      ("python-h5py" ,python-h5py)
>+      ("python-lxml" ,python-lxml)
>+      ("python-openpyxl" ,python-openpyxl)
>+      ("python-psycopg2" ,python-psycopg2)
>+      ("python-pyyaml" ,python-pyyaml)
>+      ("python-requests" ,python-requests)
>+      ("python-xlrd" ,python-xlrd)
>+      ("python-pandas" ,python-pandas)))

Are these all native inputs (required to make the build system run)? I 
expect that some of them could be moved to inputs (run-tine dependencies).

>+   (synopsis "Visidata: A terminal spreadsheet multitool for discovering and arranging data")

guix lint reports that the synopsis should not start with the package 
name. In this case, I think it could be changed to, "Terminal spreadsheet 
for discovering and arranging data".

>+   (description
>+    "VisiData is an interactive multitool for tabular data. It combines the

Please use two spaces between sentences in the description.

>+clarity of a spreadsheet, the efficiency of the terminal, and the power of
>+Python, into a lightweight utility which can handle millions of rows with ease.")

Nitpick: "with ease" sounds like a marketing term, and I believe we can 
leave it out here.

>+   (home-page "https://www.visidata.org/")
>+   (license gpl3)))

In other modules, we import licenses with the license: prefix, so we can 
refer to licenses as like license:gpl3. It might be nice to do that 
here for consistency as well.

The vdtui.py file is under the expat license, so we should record that 
here. With these changes the license field would become something like:

(license (list license:gpl3
                license:expat)) ; visidata/vdtui.py

Thanks and all the best,
Jack