Most popular programming languages 1965-2019 (visualised)

[Ola Fosheim Grøstad]

On Friday, 11 October 2019 at 08:06:02 UTC, Ola Fosheim Grøstad 
wrote:
> Keep in mind that 1986 was the heyday of 8-bit computers, low 
> on memory and diskette for storage in a blooming small business

Expanding on this so younger people understand the difference 
between the 80s and now.

1981-1986 were a period where personal computing became available 
and small businesses were able to purchase computers (with or 
without a tiny hard drive). This is kinda the 8/16 bit computing 
era. Many of these computers shipped with BASIC available. In the 
beginning when there was little software available people could 
write applications in BASIC and sell them as commercial products. 
As the market matured competition got harder and BASIC was no 
longer an option for commercial products. In the mid 80s there 
were already thousands software packages available for the IBM 
PC, and an unknown large number of similar magnitude for 8-bit 
home computers. Low memory footprint meant that programs were 
short and  focused on a limited task and that developers could 
ship new applications after only a few months of work. On 8-bit 
home computers, many of the early applications were written in 
BASIC, then a mix of BASIC and machine language and as the 
competition got harder the best apps/games were written in pure 
machine language to get most out of very limited hardware. 
Embedded programming was also typically done in machine language.

The threshold for starting up a small software company was now 
much lower than for the big mainframes... So a lot of programs 
were written, on cheap computers, using very crude tools. Some 
small developers would consider a macro assembler a luxury item...

The old computing manufactures completely missed this boat (most 
of them) and that left them in the dust. They relied on expensive 
hardware, expensive software, expensive manpower, high margins, 
small volume, large profits. So they viewed the low margin, high 
volume, small computer market as something completely separate 
and somewhat insignificant, and thus "surveys" prior to 1990 are 
likely to see this as the serious computing market that is 
completely separate from the personal computer market.

This didn't go well, IBM evaporated, SUN died, SGI died, DEC 
evaporated and so on.

1987-1994 could be viewed as the 16/32 bit era where non-GC high 
level programming took off also in the personal computing 
space... From 1995 onwards more memory was available and GC-high 
level programming and web-apps starts to dominate... This is 
where D belongs.


Anyway, measuring language popularity is problematic. Programmers 
do not necessarily like the language they have to use at work, 
and don't necessarily use the same language at home.  (We can 
assume that people who use D do so because the want to use it. 
Not so for Ada, which was known to be met with resistance and was 
most likely adopted primarily to get government contracts.)

Also, far more software is written in Java, JavaScript and PhP 
than can be seen on github. Many deployments on the web are 
closed source deployments.

So... these kinds of "I have statistics" are not as objective as 
they may seem. Getting conclusive results from quantitative 
analysis requires a very keen mindset and a lot attention to 
details and context.  There is no certainty in large numbers... 
although people find large datasets convincing. A dangerous 
fallacy...

Want to do data analysis?

1. Find high quality data.
2. Then get high quantity.
3. Then limit your findings based on a solid understanding of the 
contexts of the data-collection.